Children's Health Defense v. Meta Platforms, Inc.

                 FOR PUBLICATION

   UNITED STATES COURT OF APPEALS
        FOR THE NINTH CIRCUIT

CHILDREN’S HEALTH DEFENSE,                No. 21-16210
a Georgia non-profit organization,
                                             D.C. No.
              Plaintiff-Appellant,       3:20-cv-05787-SI
  v.

META PLATFORMS, INC., a                     OPINION
Delaware corporation; MARK
ZUCKERBERG, a California resident;
THE POYNTER INSTITUTE FOR
MEDIA STUDIES, INC., a Florida
corporation; SCIENCE FEEDBACK,
a French corporation,

              Defendants-Appellees.

       Appeal from the United States District Court
         for the Northern District of California
         Susan Illston, District Judge, Presiding

          Argued and Submitted May 17, 2022
                 Pasadena, California

                  Filed August 9, 2024
2          CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


      Before: Eric D. Miller and Daniel P. Collins, Circuit
        Judges, and Edward R. Korman, * District Judge.

                    Opinion by Judge Miller;
     Partial Concurrence and Partial Dissent by Judge Collins


                          SUMMARY **


                First Amendment/Social Media

    The panel affirmed the district court’s dismissal of a
complaint brought by the nonprofit advocacy organization
Children’s Health Defense (CHD) against Meta Platforms,
Mark Zuckerberg, and others challenging Meta’s policy of
censoring Facebook posts conveying what CHD describes as
accurate information challenging current government
orthodoxy on vaccine safety and efficacy.
    The panel noted that although Meta is a private
corporation, in certain exceptional circumstances, a private
party will be treated as a state actor for constitutional
purposes. To do so, the private party must meet two distinct
requirements: (1) the “state policy” requirement, which is
satisfied when a private institution enforces a state-imposed
rule instead of the terms of its own rules; and (2) the “state
actor” requirement, which can be met by showing, among


*
 The Honorable Edward R. Korman, United States District Judge for the
Eastern District of New York, sitting by designation.
**
  This summary constitutes no part of the opinion of the court. It has
been prepared by court staff for the convenience of the reader.
       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.       3


other things, willful participation in joint activity with the
government or government coercion.
    The panel held that CHD failed to meet the first
requirement for state action because the source of CHD’s
alleged harm was Meta’s own policy of censoring, not any
provision of federal law. The evidence suggested that Meta
had independent incentives to moderate content and
exercised its own judgment in so doing. Moreover, CHD
failed to allege any facts that would suggest an agreement
between the government and Meta that required Meta to take
a particular action in response to misinformation about
vaccines or that the government coerced Meta into
implementing a specific policy.
    The panel held that CHD’s inability to establish state
action was fatal to all of its First Amendment claims—for
damages under Bivens, for declaratory relief, and for an
injunction. To the extent that CHD argued on appeal that
Meta’s disabling of its donation button was a “taking” under
the Fifth Amendment, that claim failed for the same reason.
The panel further rejected CHD’s claim that the warning
label and fact-checks Meta placed on its posts violated the
Lanham Act, as well as CHD’s civil RICO claim.
    Concurring in part, concurring in the judgment in part,
and dissenting in part, Judge Collins stated that CHD could
plausibly allege a First Amendment claim for injunctive
relief against Meta and he therefore dissented from the
majority’s contrary conclusion. However, he agreed that all
of CHD’s other claims were properly dismissed, and he
therefore concurred in the judgment as to those remaining
claims and in Parts III, IV, and V of the majority opinion. In
Judge Collins’s view, CHD can adequately plead state action
under the test articulated in Skinner v. Railway Labor
4      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


Executives Ass’n, 

 (1989). Judge Collins would
hold that given all the circumstances, Meta’s interactions
with the Government with respect to the suppression of
specific categories of vaccine-related speech, and in
particular the speech of CHD and its founder and chairman,
Robert F. Kennedy, Jr., sufficed to implicate the First
Amendment. Because CHD could amend its complaint in a
manner that states a cause of action for injunctive and
declaratory relief, he would reverse the district court’s
judgment in favor of Meta to the extent it held to the
contrary.



                       COUNSEL

Jed Rubenfeld (argued), Yale Law School, New Haven,
Connecticut; Robert F. Kennedy Jr. (argued) and Mary S.
Holland, Children’s Health Defense, Peachtree City,
Georgia; Roger I. Teich, Roger Teich, San Francisco,
California; for Plaintiff-Appellant.
Sonal N. Mehta (argued), Wilmer Cutler Pickering Hale and
Dorr LLP, Palo Alto, California; Ari Holtzblatt, Molly M.
Jennings, Allison Schultz, and Spencer Todd, Wilmer Cutler
Pickering Hale and Dorr LLP, Washington, D.C.; Mark R.
Caramanica (argued), Daniela Abratt-Cohen, and Carol J.
LoCicero, Thomas & LoCicero PL, Tampa, Florida;
Elizabeth H. Baldridge, Jenner & Block LLP, Los Angeles,
California; Kevin L. Vick, Jassy Vick Carolan LLP, Los
Angeles, California; for Defendants-Appellees.
John W. Whitehead, The Rutherford Institute,
Charlottesville, Virginia; for Amicus Curiae Rutherford
Institute.
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.       5


                         OPINION

MILLER, Circuit Judge:

    Children’s Health Defense (CHD) is a nonprofit
advocacy organization dedicated to educating the public
about what it sees as the dangers of vaccines. The
organization regularly shares articles and videos on its
Facebook page, but since 2019, Meta Platforms, Inc., the
operator of Facebook, has restricted CHD’s ability to do so,
including by adding warning labels to alert users that, in
Meta’s view, the information that CHD shares is not
accurate.
    Believing that Meta was censoring its speech at the
direction of the federal government, CHD brought this
action against Meta; Mark Zuckerberg, Meta’s CEO; and the
Poynter Institute and Science Feedback, both of which
contract with Meta to evaluate the accuracy of some
Facebook content. It asserted claims under the First and Fifth
Amendments as well as the Lanham Act, 

(a), and the Racketeer Influenced and Corrupt
Organizations Act (RICO), 

. The district
court dismissed the complaint. We affirm.
                               I
    Because this is an appeal from an order granting a motion
to dismiss, we assume the truth of the facts alleged in the
operative complaint—here, CHD’s second amended
complaint. Ellis v. Salt River Project Agric. Improvement &
Power Dist., 

, 1266 (9th Cir. 2022). After filing
that complaint, CHD moved to “supplement” it with
additional allegations, filed a motion for judicial notice that
contained further allegations, and then moved to “further
6       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


supplement” the complaint. The district court denied CHD
leave to amend the complaint but considered the allegations
within CHD’s motions “as a further proffer of how CHD
would amend the complaint if given leave to do so.” We
have likewise considered those allegations, and they are
reflected in the description of the facts set out below.
    CHD describes itself as an organization that seeks “to
provide the public with timely and accurate vaccine and 5G
and wireless technology safety information.” To that end,
CHD publishes articles and opinion pieces on its eponymous
website and on its Facebook page. Those writings often
describe purported links between vaccinations and various
illnesses. CHD has posted articles that it claims show that
“[u]nvaccinated kids are healthier” than their vaccinated
counterparts. Sometimes, CHD posts messages from its
founder, Robert F. Kennedy, Jr., in which he criticizes Dr.
Anthony Fauci and Bill Gates and their efforts to encourage
vaccinations.
    Although public discussion of vaccines has taken on a
new dimension as a result of the COVID-19 pandemic, some
lawmakers have expressed concern about the proliferation of
“vaccine misinformation” on social media platforms for
several years. In February 2019, Representative Adam
Schiff of California sent a public letter to Zuckerberg, asking
(1) whether “medically inaccurate information” violated
Facebook’s terms of service; (2) what steps Facebook took
to address “misinformation related to vaccines” and whether
it planned to take additional steps; (3) whether Facebook
allowed anti-vaccine activists and organizations to advertise
on its platform; and (4) what steps Facebook took to prevent
its algorithm from recommending anti-vaccine content to
users. After COVID-19 vaccines became widely available,
some lawmakers expressed renewed concern that social
       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.       7


media companies like Meta were not doing enough to slow
the spread of false information about the virus and vaccines.
For example, Senator Amy Klobuchar of Minnesota wrote
to Zuckerberg, stating that Facebook’s “policies must be
strictly enforced to limit users’ exposure to misinformation”
and urging him to “take action against people that are
spreading content that can harm the health of Americans.”
    For its part, Meta announced in early 2019 that it had
begun to “tackle vaccine misinformation” on Facebook by
making that content less prominent in search results,
rejecting ads that included it, and “exploring ways to share
educational information about vaccines when people come
across misinformation on this topic.” It promised to “take
action” against posts that shared “verifiable vaccine hoaxes,”
as defined by the World Health Organization (WHO) and the
U.S. Centers for Disease Control and Prevention (CDC).
   After those policies were announced, CHD noticed
changes to the functionality and appearance of its Facebook
page. A banner was placed at the top of its page, with a
message that read:

       This Page posts about vaccines
       When it comes to health, everyone wants
       reliable, up-to-date information. The Centers
       for Disease Control (CDC) has information
       that can help answer questions you may have
       about vaccines.
       Go to CDC.gov

   Around the same time, Meta began flagging CHD’s
posts as containing factual inaccuracies. To identify content
posted on Facebook that it considers inaccurate, Meta
8      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


contracts with the Poynter Institute (which operates a
website known as “PolitiFact”) and Science Feedback.
Specifically, Meta directs those services to review and
classify content that its algorithms have identified as
potentially containing “misinformation.” If the reviewers
determine that the content contains false or misleading
information, it may appear under a grey overlay that informs
readers that the post has been labeled false and refers them
to a link so that they can “See Why.” The link leads users to
a new window that contains a short explanation of the
classification—for example, that independent fact-checkers
have determined that the information shared in the post is
“factually inaccurate.” The contents of the flagged post
remain accessible, but visitors must click a slightly less
prominent link in order to view it. If Meta determines that
the post violates Facebook’s Community Standards, it may
be removed entirely.
     After identifying repeated factual inaccuracies in CHD’s
posts, Meta deactivated the “donate” button on CHD’s page,
telling the group that it had violated Facebook’s “fundraising
terms and conditions.” Before this happened, CHD had
received more than $40,000 in donations through its
Facebook page in 2019. Meta also prohibited CHD,
Kennedy, and an agency employed by the two from
purchasing advertisements on Facebook because, it said,
CHD had “repeatedly posted content that has been disputed
by third-party fact-checkers [for] promoting false content.”
    As part of its response to the COVID-19 pandemic, Meta
took further action against CHD. It updated Facebook’s
policies to prohibit users from sharing any “claims that
COVID-19 vaccines are not effective in preventing COVID-
19,” and it created a “Coronavirus (COVID-19) Information
Center,” which links to the CDC’s website and other
       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.      9


“leading health organizations” for information on the
pandemic. Meta then began displaying messages to CHD’s
followers encouraging them to unsubscribe from its posts
and referring them to the WHO for facts about COVID-19.
     CHD alleges that Meta has also limited the visibility of
its content using processes known as “shadow-banning” and
“sandboxing.” With shadow-banning, Meta allows a post to
remain visible to the poster, and in some cases the poster’s
Facebook “friends,” while hiding the post from other users.
With sandboxing, Meta shows CHD’s posts about vaccines
to like-minded users but not to those who do not already
share its views. CHD says that, as a result, traffic to its
website from its Facebook page has declined significantly.
Although CHD once had the ability to dispute the actions
Meta took with respect to its page, Meta disabled that
functionality, and it has not been restored.
    In August 2020, CHD brought this action in the Northern
District of California. It alleged that Meta, Zuckerberg, the
Poynter Institute, and Science Feedback were working in
concert with or, alternatively, under compulsion from the
federal government to censor CHD’s speech, in violation of
the First Amendment, and to deprive it of its property right
to fundraise on Facebook, in violation of the Fifth
Amendment. Based on those alleged constitutional
violations, CHD sought damages under Bivens v. Six
Unknown Named Agents of Federal Bureau of Narcotics,

 (1971). It also sought injunctive and
declaratory relief. CHD further claimed that the defendants
violated the Lanham Act by labeling its posts as false, and
that the defendants imposed those labels as part of a
fraudulent scheme to divert donations away from CHD for
the benefit of Meta’s fact-checkers, in violation of RICO.
10     CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


CHD sought money damages as well as injunctive and
declaratory relief for those claims too.
    Meta, Zuckerberg, and the Poynter Institute moved to
dismiss, and the district court dismissed the complaint
without leave to amend. The court held that CHD’s
constitutional claims failed because “CHD has not alleged
that the challenged acts constitute federal action.”
Specifically, the court determined that “general statements
by the CDC and Zuckerberg about ‘working together’ to
reduce the spread of health or vaccine misinformation, or to
promote universal vaccination do not show that the
government was a ‘joint participant in the challenged
activity.’” It emphasized that CHD had not “alleged that the
government was actually involved in the decisions to label
CHD’s posts as ‘false’ or ‘misleading,’ the decision to put
the warning label on CHD’s Facebook page, or the decisions
to ‘demonetize’ or ‘shadow-ban.’” The court further
concluded that “CHD has not alleged facts showing
government coercion sufficient to deem Facebook or
Zuckerberg a federal actor.”
    The district court also rejected the Lanham Act claim. It
explained that “the warning label and fact-checks are not
disparaging CHD’s ‘goods or services,’ nor are they
promoting the ‘goods or services’ of Facebook, the CDC, or
the fact-checking organizations such as Poynter.” For those
reasons, the court concluded that “CHD’s alleged injuries
are not within the Lanham Act’s ‘zone of interests’ and that
the warning label and fact-checks are not ‘commercial
advertising or promotion’” within the scope of the statute.
    The district court rejected the RICO claim because CHD
had not established a predicate act of wire fraud. The court
stated that “CHD’s allegations . . . do not constitute wire
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.       11


fraud because CHD has not alleged any facts showing that
defendants engaged in a fraudulent scheme to obtain money
or property from Facebook visitors to CHD’s page.”
    Science Feedback is a French nonprofit organization,
and CHD was apparently unable to serve it with process. As
a result, the district court dismissed all claims against
Science Feedback without prejudice.
    CHD appeals. We review the district court’s grant of a
motion to dismiss de novo. Wells Fargo Bank, N.A. v.
Mahogany Meadows Ave. Tr., 

, 1213 (9th Cir.
2020). “To survive a motion to dismiss, a complaint must
contain sufficient factual matter, accepted as true, to ‘state a
claim to relief that is plausible on its face.’” Ashcroft v.
Iqbal, 

 (2009) (quoting Bell Atl. Corp. v.
Twombly, 

 (2007)).
                               II
    The First Amendment provides that “Congress shall
make no law . . . abridging the freedom of speech.” U.S.
Const. amend. I. Within its scope, the First Amendment
provides robust protection for free speech. But it has an
important limitation: It “prohibits only governmental
abridgment of speech” and “does not prohibit private
abridgment of speech.” Manhattan Cmty. Access Corp. v.
Halleck, 

, 808 (2019); see Prager Univ. v.
Google LLC, 

, 996 (9th Cir. 2020).
     That limitation is itself an important protection for
liberty. If the First Amendment were applied to private
actors, it would mean, for example, that a newspaper would
be unable to choose to print the work of only those writers
whose views were consistent with its editorial positions, and
it could instead be forced by the federal courts to open itself
12      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


to all writers on a nondiscriminatory basis. See Miami
Herald Publ’g Co. v. Tornillo, 

, 254–58 (1974).
“By enforcing [the] constitutional boundary between the
governmental and the private, the state-action doctrine”
developed by the Supreme Court to distinguish government
from private action “protects a robust sphere of individual
liberty.” Halleck, 587 U.S. at 808; accord Lugar v.
Edmondson Oil Co., 

 (1982) (“Careful
adherence to the ‘state action’ requirement preserves an area
of individual freedom by limiting the reach of federal law
and federal judicial power.”).
    To begin by stating the obvious, Meta, the owner of
Facebook, is a private corporation, not a government agency.
Although that fact is highly relevant here, it does not quite
end our inquiry because, in certain “exceptional cases,” a
private party “will be treated as a state actor for
constitutional purposes.” O’Handley v. Weber, 

, 1155–56 (9th Cir. 2023). The private party must meet
two distinct requirements: (1) the “state policy” requirement
and (2) the “state actor” requirement. Wright v. Service
Emps. Int’l Union Loc. 503, 

, 1121 (9th Cir.
2022); see Lugar, 

; O’Handley, 62 F.4th at
1156.
    To satisfy the state policy requirement, the alleged
constitutional deprivation must result from “the exercise of
some right or privilege created by the State” or “a rule of
conduct imposed by the State or by a person for whom the
State is responsible.” Lugar, 

. To satisfy the
state actor requirement, the party must “fairly be said to be a
state actor,” 

 which requires that it meet one of four tests:
(1) the private actor performs a traditionally public function,
Halleck, 587 U.S. at 804; (2) the private actor is a “willful
participant in joint activity” with the government, Lugar,
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.      13

 (quoting Adickes v. S. H. Kress & Co., 

 (1970)); (3) the government compels or
encourages the private actor to take a particular action, Blum
v. Yaretsky, 

 (1982); or (4) there is a
“sufficiently close nexus” between the government and the
challenged action, Jackson v. Metropolitan Edison Co., 

 (1974).
    This test for state action “ensures that not all private
parties ‘face constitutional litigation whenever they seek to
rely on some state rule governing their interactions with the
community surrounding them.’” Collins v. Womancare, 

 (9th Cir. 1989) (quoting Lugar, 

). At bottom, both components of the test ask us to
evaluate whether the nature of the relationship between the
private party and the government is such that “the alleged
infringement of federal rights is fairly attributable to the
[government].” Pasadena Republican Club v. Western Just.
Ctr., 

, 1167 (9th Cir. 2021) (alteration in
original) (quoting Sutton v. Providence St. Joseph Med. Ctr.,

 (9th Cir. 1999)). In other words, a plaintiff
must allege facts supporting an inference that the
government “is responsible for the specific conduct of which
the plaintiff complains.” Ohno v. Yasuma, 

(9th Cir. 2013) (quoting Blum, 

).
                              A
    We first look to whether the “‘source of the alleged
constitutional harm’” is “a state statute or policy.” Belgau v.
Inslee, 

, 947 (9th Cir. 2020) (quoting Ohno, 

). This requirement is satisfied when a private
institution “enforce[s] a state-imposed rule” instead of “the
terms of its own rules.” O’Handley, 62 F.4th at 1156.
14     CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


    CHD’s state-action theory fails at this threshold step. We
begin our analysis by identifying the “specific conduct of
which the plaintiff complains.” Wright, 48 F.4th at 1122
(quoting American Mfrs. Mut. Ins. Co. v. Sullivan, 

 (1999)). CHD challenges Meta’s “policy of
censoring” posts conveying what it describes as “accurate
information . . . challenging current government orthodoxy
on . . . vaccine safety and efficacy.” But “the source of the
alleged . . . harm,” Ohno, 

, is Meta’s own
“policy of censoring,” not any provision of federal law. The
closest CHD comes to alleging a federal “rule of conduct” is
the CDC’s identification of “vaccine misinformation” and
“vaccine hesitancy” as top priorities in 2019. But as we
explain in more detail below, those statements fall far short
of suggesting any actionable federal “rule” that Meta was
required to follow. And CHD does not allege that any
specific actions Meta took on its platforms were traceable to
those generalized federal concerns about vaccine
misinformation.
     In O’Handley, we rejected a claim similar to CHD’s
asserted by a Twitter user who objected to Twitter’s decision
to limit access to his tweets and suspend his account. See 62
F.4th at 1156. The user alleged that Twitter’s actions were
prompted by a message from the California Secretary of
State identifying one of the user’s tweets as spreading
election-related “disinformation.” Id. at 1154. But because
“the company acted under the terms of its own rules, not
under any provision of California law,” we rejected the
argument that Twitter “ceded control over [its] content-
moderation decisions to the State and thereby became the
government’s private enforcer[].” Id. The same is true here.
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.        15


                               B
     CHD’s failure to satisfy the first part of the test is fatal
to its state action claim. See Lindke v. Freed, 

 (2024); but see O’Handley, 62 F.4th at 1157
(noting that our cases “have not been entirely consistent on
this point”). Even so, CHD also fails under the second part.
As we have explained, the Supreme Court has identified four
tests for when a private party “may fairly be said to be a state
actor”: (1) the public function test, (2) the joint action test,
(3) the state compulsion test, and (4) the nexus test. Lugar,

.
    CHD invokes two of those theories of state action as well
as a hybrid of the two. First, it argues that Meta and the
federal government agreed to a joint course of action that
deprived CHD of its constitutional rights. Second, it argues
that Meta deprived it of its constitutional rights because
government actors pressured Meta into doing so. Third, it
argues that the “convergence” of “joint action” and
“pressure,” as well as the “immunity” Meta enjoys under 

, make its allegations that the government used
Meta to censor disfavored speech all the more plausible.
CHD cannot prevail on any of these theories.
                               1
    The joint action test asks “whether the government has
so far insinuated itself into a position of interdependence
with a private entity that the private entity must be
recognized as a joint participant in the challenged activity.”
Pasadena Republican Club, 985 F.3d at 1167 (quoting
Brunette v. Humane Soc’y of Ventura Cnty., 

 (9th Cir. 2002)). Our cases require a plaintiff to plead
facts that give rise to an inference that the private entity’s
16      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


“particular actions are ‘inextricably intertwined’ with those
of the government.” 

 (quoting Brunette, 

).
    Crucially, it is not enough to show an agreement to do
something; the private party and the government must also
have agreed on what the something is. “The generalized
allegation of a wink and a nod understanding . . . does not
amount to an agreement or a conspiracy to violate [the
plaintiff’s] rights in particular.” Brunette, 

.
Thus, a plaintiff must show some specificity to the
understanding between the private actor and the government.
See Dennis v. Sparks, 

, 25–28 (1980) (agreement
between litigants and judge to issue an illegal injunction
preventing production on oil leases); Adickes, 

 (agreement between store employee and police officer
to arrest the plaintiff); Swift v. Lewis, 

, 731–32
& n.2 (9th Cir. 1990) (agreement between prison and
contractor to remove petitioner’s religious classification),
superseded by statute on other grounds, 42 U.S.C. § 2000cc-
1(a); Howerton v. Gabica, 

, 384–85 (9th Cir.
1983) (agreement between landlord and police officer to
evict plaintiffs).
    CHD has not done so. In an effort to show an agreement,
CHD points to various statements from Meta and
government officials, but they suffer from a critical lack of
specificity. For example, CHD highlights the CDC’s
statement that it has “engaged” social media companies to
“contain the spread of misinformation.” That could mean
many different things, thanks to ambiguity in both the verb
(“Containing” misinformation by removing it entirely? By
making it less prominent on the site? By leaving it as is but
countering it with different information?) and its object
(What counts as “misinformation”?). The “generic
promotion of a public purpose” falls far short of establishing
       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.      17


that Meta’s “particular actions are inextricably intertwined
with those of the government.” Pasadena Republican Club,
985 F.3d at 1167, 1170 (internal quotation marks omitted).
Without plausible allegations of an agreement to take
specific action, we cannot say that Meta’s conduct is fairly
attributable to the government.
    CHD asks us to infer a more specific agreement among
Meta, the Biden Administration, the CDC, and the WHO, in
which Meta took direction from those entities about what
content to censor. But the facts that CHD alleges do not
make that inference plausible in light of the obvious
alternative—that the government hoped Meta would
cooperate because it has a similar view about the safety and
efficacy of vaccines. See Twombly, 550 U.S. at 556–57.
Links between a social media company’s communications
with the government and its decisions about what content to
permit “must be evaluated in light of the platform’s
independent incentives to moderate content.” Murthy v.
Missouri, 

, 1988 (2024) (rejecting similar
claims that government officials and agencies pressured
platforms to unconstitutionally suppress COVID-19
misinformation). Statements that government officials
“engaged” with social media companies to ensure that those
companies “understand the importance of misinformation
and disinformation and how they can get rid of it quickly”
are consistent with the explanation of parallel objectives and
do not show the specific agreement that CHD suggests. As
for the WHO, it is an intergovernmental agency, not part of
the federal government, so its meeting with Meta in which it
“discussed” Meta’s “role in spreading ‘lifesaving health
information’” is irrelevant to the state-action inquiry.
    In a belated attempt to bolster its theory, CHD asks us to
take judicial notice of various documents showing that the
18     CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


government works with social media companies to educate
them about what it considers to be misinformation on their
platforms. “[W]e rarely take judicial notice of facts
presented for the first time on appeal.” Reina-Rodriguez v.
United States, 

 (9th Cir. 2011). We do
not think it is appropriate to do so here, especially because
the facts CHD presents are not free from “reasonable
dispute.” Fed. R. Evid. 201(b). To be sure, it would be proper
to take judicial notice of the fact that the documents exist.
Cf. Lee v. City of Los Angeles, 

 (9th Cir.
2001). But CHD’s allegations rely on the substance of the
documents and what the statements in them establish.
Because those statements are “subject to varying
interpretations,” they cannot qualify for judicial notice.
Reina-Rodriguez, 

.
    In any event, even if we were to consider the documents,
they do not make it any more plausible that Meta has taken
any specific action on the government’s say-so. To the
contrary, they indicate that Meta and the government have
regularly disagreed about what policies to implement and
how to enforce them. See Murthy, 144 S. Ct. at 1987
(highlighting evidence “that White House officials had
flagged content that did not violate company policy”). Even
if Meta has removed or restricted some of the content of
which the government disapproves, the evidence suggests
that Meta “had independent incentives to moderate content
and . . . exercised [its] own judgment” in so doing. Id.
    That the government submitted requests for removal of
specific content through a “portal” Meta created to facilitate
such communication does not give rise to a plausible
inference of joint action. Exactly the same was true in
O’Handley, where Twitter had created a “Partner Support
Portal” through which the government flagged posts to
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.      19


which it objected. 62 F.4th at 1160. Meta was entitled to
encourage such input from the government as long as “the
company’s employees decided how to utilize this
information based on their own reading of the flagged
posts.” Id. It does not become an agent of the government
just because it decides that the CDC sometimes has a point.
     The circumstantial evidence that CHD proffers does not
nudge its claims into the realm of plausibility either. CHD
alleges, for example, that when Meta deactivated the
“donate” button on CHD’s page in May 2019, it must have
done so because of the letter Representative Schiff sent to
Meta that February. In the letter, Schiff expressed concern
that misleading or incorrect information about vaccines was
leading to a decline in vaccine uptake. He asked Meta to
explain how it dealt with such content on its platform, and
he “encourage[d] [Meta] to consider . . . additional steps.” It
is simply not reasonable to infer from those two events that
Meta “takes direction from the federal government about
what COVID-related speech to censor,” as CHD would have
it.
    Failing to allege a plausible agreement between Meta
and the government, CHD seeks to fill the gap by arguing
that the CDC supplied Meta with a “standard of decision” by
which allegedly unconstitutional actions were taken.
Pointing to statements from Zuckerberg announcing that
Meta defers to the CDC for “authoritative information,”
CHD asserts that the algorithms Meta implemented to flag
misinformation apply “agreed-to, government-provided
standards of decision.”
   CHD invokes Mathis v. Pacific Gas & Electric Co.
(Mathis I), in which we allowed a Bivens action to proceed
against PG&E, a public utility company, because we
20      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


concluded that a former employee of a PG&E contractor had
plausibly alleged that PG&E denied him access to a nuclear
power plant “on the basis of some rule of decision for which
the State is responsible.” 

, 1432 (9th Cir.
1989) (quoting Rendell-Baker v. Kohn, 

(1982) (White, J., concurring in the judgments)).
Specifically, Mathis had been denied access to the plant and
then fired because he was suspected of selling or using
illegal drugs. Id. at 1430, 1432–33. He alleged that PG&E
denied him access because the Nuclear Regulatory
Commission had pressured and encouraged PG&E to adopt
a policy of excluding from nuclear power facilities anyone
who sold or used drugs. See id. at 1433; see also Mathis v.
Pacific Gas & Elec. Co. (Mathis II), 

, 502–03
(9th Cir. 1996) (requiring a showing that a “standard
compelled” a certain decision). If that allegation were true,
we reasoned, it would establish that PG&E’s decision to
exclude Mathis was fairly attributable to the government.
Mathis I, 891 F.2d at 1434.
    The allegations here are far different from those in
Mathis I. Mathis alleged the existence of an informal
government policy that required the utility to take a specific
action in response to certain events. CHD has alleged that
Meta banned “vaccine misinformation” and that it defers to
the CDC for “authoritative information” on that topic. It has
failed, however, to allege any facts that would allow us to
infer an agreement between the government and Meta that
required Meta to take a particular action in response to
misinformation about vaccines. Further, as we have already
explained, “misinformation” is far too amorphous a concept
to serve as the type of “standard” contemplated by Mathis I.
See 891 F.2d at 1433–34. And without a standard that can
plausibly be said to require a specific outcome, it is not fair
       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.     21


to say that Meta’s “choice must in law be deemed to be that
of the State.” Blum, 

.
    Finally, CHD argues that financial benefits flowing from
Meta to the government support an inference that Meta’s
conduct constitutes state action. In so doing, it invokes the
Supreme Court’s observation that a plaintiff may
“sometimes” be able to prove government responsibility for
a nominally private action if the government “knowingly
accepts the benefits derived from unconstitutional
behavior.” National Collegiate Athletic Ass’n v. Tarkanian,

 (1988). The putative “benefits” here are
$35 million Zuckerberg and Meta have donated to the CDC
Foundation and the “millions of dollars in free advertising”
and reputational benefits Meta has given the CDC. But those
benefits are not directly tied to the specific action being
challenged in this case: restricting CHD’s Facebook posts.
CHD therefore has not alleged the kind of “significant
financial integration” that we have found probative in
determining whether the joint-action test is satisfied.
Pasadena Republican Club, 985 F.3d at 1168 (quoting
Brunette, 

). To the contrary, Zuckerberg and
Meta’s donations to the CDC Foundation make the innocent
alternative—that Meta adopted the policy it did simply
because Zuckerberg and Meta share the government’s view
that vaccines are safe and effective—all the more plausible.
    We acknowledge that there is a degree of uncertainty in
determining how specific the details of an agreement must
be before a plaintiff can be said to have plausibly alleged
joint action. The Supreme Court has remarked that the state-
action inquiry is a “matter of normative judgment” whose
“criteria lack rigid simplicity,” so some uncertainty is
inherent in the doctrine. Brentwood Acad. v. Tennessee
Secondary Sch. Athletic Ass’n, 

 (2001).
22      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


But wherever the line may be, CHD is far from it. Rather
than pleading facts that allow us to infer that the government
and Meta agreed to censor speech on Facebook, CHD has
alleged that the government hoped Meta would cooperate in
its efforts to promote the safety and efficacy of vaccines.
    Meta has a First Amendment right to use its platform to
promote views it finds congenial and to refrain from
promoting views it finds distasteful: “Like . . . editors, cable
operators, and parade organizers,” social media companies
make “choices about whether—and, if so, how—to convey
posts having a certain content or viewpoint” that “rest on a
set of beliefs about which messages are appropriate and
which are not.” Moody v. NetChoice, LLC, 

,
2405 (2024). Even though companies like Meta “happily
convey the lion’s share of posts submitted to them,” it
remains “as much an editorial choice to convey all speech
except in select categories as to convey only speech within
them.” 

. “When the platforms use their Standards
and Guidelines to decide which third-party content [their]
feeds will display, or how the display will be ordered and
organized, they are making expressive choices. And because
that is true, they receive First Amendment protection.” 

    Meta evidently believes that vaccines are safe and
effective and that their use should be encouraged. It does not
lose the right to promote those views simply because they
happen to be shared by the government.
                               2
    A private party may also be considered a state actor if it
has acted because the government coerced or compelled it to
do so. Under the coercion test, the government must have
“exercised coercive power or . . . provided such significant
encouragement, either overt or covert, that the choice must
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.      23


in law be deemed to be that of the State.” Blum, 

. The government’s “[m]ere approval of” private
initiatives, however, “is not sufficient to justify holding the
State responsible for those initiatives.” 

 at 1004–05.
Instead, the government must “convey a threat of adverse
government action,” National Rifle Ass’n of Am. v. Vullo,

 (2024), or otherwise impose incentives
that “overwhelm” and “essentially compel” the party to
comply with its requests, O’Handley, 62 F.4th at 1158.
    At the outset, we note that there is reason to doubt that a
purely private actor like Meta, which was the victim of the
alleged coercion, would be the appropriate defendant, rather
than the government officials responsible for the coercion.
In Sutton v. Providence St. Joseph Medical Center, for
example, we said that “only the state actor, and not the
private party, should be held liable for the constitutional
violation that resulted from the state compulsion.” 

 (quoting Barbara Rook Snyder, Private Motivation,
State Action and the Allocation of Responsibility for
Fourteenth Amendment Violations, 

,
1067 (1990)). But see generally Carlin Commc’ns, Inc. v.
Mountain States Tel. & Tel. Co., 

 (9th Cir.
1987). We need not resolve that question here because CHD
has not adequately pleaded facts supporting a coercion
theory of state action.
    CHD’s theory of coercion turns on statements made by
lawmakers threatening to hold social media companies
“accountable” for failing to police “misinformation” on their
platforms. Those statements do not meet the standard we
have articulated for finding state action. Here again, the key
case arises from the Mathis litigation, this time Mathis II. In
his second appeal, Mathis argued that he had proved his
claim that PG&E excluded him from the power plant
24      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


because it applied a “standard of decision” imposed on the
utility by the government. Mathis II, 75 F.3d at 502 (quoting
Mathis I, 891 F.2d at 1434). But rather than demonstrating
that the government pressured the utility into adopting a
specific policy requiring his exclusion, Mathis showed only
that PG&E “was aware of a generalized federal concern with
drug use at nuclear power plants” and that PG&E “was
looking to score Brownie points” with the government by
adopting a policy to address that concern. Id. Mathis argued
that he came “close enough” to proving coercion because his
evidence gave rise to the inference that PG&E implemented
a drug-use policy to “allay [the government’s] concerns.” Id.
at 503. But we rejected that argument. We explained that
Mathis “asks us to hold that regulatory interest in a problem
transforms any subsequent private efforts to address the
problem (even those expressly designed to obviate the need
for regulation) into state action.” Id. We refused to do so,
adding that “[i]f the government is considering regulation,
affected private parties can try to convince it there’s no need
to regulate without thereby transforming themselves into the
state’s agents.” Id.
     CHD has not alleged facts that allow us to infer that the
government coerced Meta into implementing a specific
policy. Instead, it cites statements by Members of Congress
criticizing social media companies for allowing
“misinformation” to spread on their platforms and urging
them to combat such content because the government would
hold them “accountable” if they did not. Like the
“generalized federal concern[s]” in Mathis II, those
statements do not establish coercion because they do not
support the inference that the government pressured Meta
into taking any specific action with respect to speech about
vaccines. Mathis II, 75 F.3d at 502. Indeed, some of the
       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.     25


statements on which CHD relies relate to alleged
misinformation more generally, such as a statement from
then-candidate Biden objecting to a Facebook ad that falsely
claimed that he blackmailed Ukrainian officials. All CHD
has pleaded is that Meta was aware of a generalized federal
concern with misinformation on social media platforms and
that Meta took steps to address that concern. See id. If Meta
implemented its policy at least in part to stave off
lawmakers’ efforts to regulate, it was allowed to do so
without turning itself into an arm of the federal government.
See id. at 503.
    CHD argues that the letters sent to Meta by Senator
Klobuchar and Representative Schiff demonstrate the
necessary coercion. In one of Klobuchar’s letters, she urged
Meta to “take action” against prominent anti-vaccine
influencers such as Kennedy. In another letter, she asked
Meta a series of questions about its handling of “vaccine-
related misinformation,” told it that transparency was
“imperative,” and said that “policies must be strictly
enforced to limit users’ exposure to misinformation.”
Schiff’s letter was along similar lines. But in contrast to
cases where courts have found coercion, the letters did not
require Meta to take any particular action and did not
threaten penalties for noncompliance. See National Rifle
Ass’n of Am., 

 (agency superintendent
promising to “ignore” insurance-law violations if insurer
“ceased underwriting NRA policies and disassociated from
gun-promotion groups”); Bantam Books, Inc. v. Sullivan,

, 62 n.5 (1963) (state commission notifying book
distributor of the names of “obscene publications” that were
“objectionable for sale” and implying that the Attorney
General would prosecute if the bookseller did not
cooperate); Carlin Commc’ns, 827 F.2d at 1295 (county
26      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


attorney threatening to prosecute if a utility did not terminate
plaintiff’s service); Backpage.com, LLC v. Dart, 

, 230–31 (7th Cir. 2015) (county sheriff demanding that
credit card companies “immediately cease and desist” from
allowing their cards to be used to purchase advertisements
on an adult website); Okwedy v. Molinari, 

,
341–42 (2d Cir. 2003) (per curiam) (city borough president
objecting to a message on a billboard designed by a media
company and directing the company to contact his counsel).
    Moreover, “[t]he power that a government official
wields . . . is relevant to the objective inquiry of whether a
reasonable person would perceive the official’s
communication as coercive.” National Rifle Ass’n of Am.,

. “[D]irect regulatory and enforcement
authority,” such as the ability to “initiate investigations and
refer cases for prosecution,” makes coercion more likely. 

. By contrast, “[a] letter from a single Senator backed
by no statutory mandate is far afield from [a] system of
‘effective state regulation’” that would suggest coercion.
Kennedy v. Warren, 

, 1210 (9th Cir. 2023)
(quoting Bantam Books, 327 U.S. at 69). Unlike “an
executive official with unilateral power that could be
wielded in an unfair way if the recipient did not acquiesce,”
a single legislator lacks “unilateral regulatory authority.” Id.
A letter from a legislator would therefore “more naturally be
viewed as relying on her persuasive authority rather than on
the coercive power of the government.” Id.
    The statements here are firmly on the constitutional side
of the sometimes “fine lines between permissible
expressions of personal opinion and implied threats to
employ coercive state power to stifle protected speech.”
Hammerhead Enters., Inc. v. Brezenoff, 

 (2d
Cir. 1983); see O’Handley, 62 F.4th at 1158 (holding that
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.      27


Twitter’s compliance with a particularized government
“request with no strings attached” was the product of the
company’s “own independent judgment”).
                              3
    CHD also advances a hybrid theory of joint action and
coercion that focuses on section 230 of the Communications
Decency Act, 

. That provision states that
“[n]o provider or user of an interactive computer service
shall be treated as the publisher or speaker of any
information provided by another information content
provider.” 

 § 230(c)(1). It also immunizes providers of
interactive computer services from civil liability for “any
action voluntarily taken in good faith to restrict access to or
availability of material that the provider or user considers to
be obscene, lewd, lascivious, filthy, excessively violent,
harassing, or otherwise objectionable, whether or not such
material is constitutionally protected.” Id. § 230(c)(2)(A).
    The immunity from liability conferred by section 230 is
undoubtedly a significant benefit to companies like Meta
that operate social media platforms. It might even be the case
that such platforms could not operate at their present scale
without section 230. But many companies rely, in one way
or another, on a favorable regulatory environment or the
goodwill of the government. If that were enough for state
action, every large government contractor would be a state
actor. But that is not the law.
   CHD seeks to analogize section 230 to the regulatory
scheme that the Supreme Court considered in Skinner v.
Railway Labor Executives’ Ass’n, but the analogy is inapt.

 (1989). That case involved Federal Railroad
Administration (FRA) regulations authorizing railroads to
conduct drug tests on employees suspected of violating
28      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


certain safety rules. 

. The FRA argued that
because the regulations merely permitted testing, but did not
require it, the tests would not constitute state action and
would not implicate the Fourth Amendment. 

. The
Court rejected that argument, reasoning that the “specific
features of the regulations” demonstrated that “the
Government did more than adopt a passive position toward
the underlying private conduct.” 

. In particular, the
regulations preempted state laws, superseded “any provision
of a collective bargaining agreement,” and prohibited a
railroad from “divest[ing] itself of” or “otherwise
compromis[ing] by contract” the ability to conduct the tests.

 (citation omitted). In addition, the regulations gave the
FRA “the right to receive certain biological samples and test
results”—the “fruits” of the searches—and mandated that
any employee who refused to undergo a test “be withdrawn
from covered service.” 

 Those features, considered
together, made the Court “unwilling to accept” the argument
that any searches would be “primarily the result of private
initiative.” 

    CHD argues that because the immunity in section 230,
like the regulatory regime in Skinner, “removed all legal
barriers” to the censorship of vaccine-related speech, Meta’s
restriction of that content should be considered state action.
See 

. But section 230 is fundamentally unlike
the regulations in Skinner. The statute is entirely passive—a
provider can leave content on its platform without worrying
that the speech of the poster will be imputed to it, or it may
choose to restrict content it considers “objectionable”
without the threat of lawsuits. Significantly, in Skinner, the
removal of “legal barriers” was just one among several
facets of the regulatory scheme that the Court cited in finding
state compulsion. 

 Under that scheme, the government
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.        29


sought to encourage railroads both to test their employees
and to share the “fruits” of those tests with the government.

 As evidence of such encouragement, the Court noted that
the government imposed on railroads a “duty to promote the
public safety” and “mandated” that they fulfill that duty by
preserving their state-conferred “authority to perform tests.”

 (citation omitted).
    Such “indices of the Government’s encouragement,
endorsement, and participation” to promote particular
private conduct are absent here. 

 at 615–16. Section 230
is just as protective of a provider’s right to maintain
“objectionable” content on its platform as it is of a provider’s
right to delete such content. The “legislative grace”
providers enjoy under Section 230 merely affords them the
ability to choose whether to suppress certain third-party
speech without risking costly litigation. By giving
companies like Meta that freedom, the government has
hardly expressed a “strong preference” for the removal of
speech critical of vaccines. 

    It would be exceptionally odd to say that the government,
through section 230, has expressed any preference at all as
to the removal of anti-vaccine speech, because the statute
was enacted years before the government was concerned
with speech related to vaccines, and the statute makes no
reference to that kind of speech. Rather, as the text of section
230(c)(2)(A) makes clear—and as the title of the statute (i.e.,
the “Communications Decency Act”) confirms—a major
concern of Congress was the ability of providers to restrict
sexually explicit content, including forms of such content
that enjoy constitutional protection. It is not difficult to find
examples of Members of Congress expressing concern about
sexually explicit but constitutionally protected content, and
many providers, including Facebook, do in fact restrict it.
30     CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


See, e.g., 141 Cong. Rec. 22,045 (1995) (statement of Rep.
Wyden) (“We are all against smut and pornography . . . .”);
id. at 22,047 (statement of Rep. Goodlatte) (“Congress has a
responsibility to help encourage the private sector to protect
our children from being exposed to obscene and indecent
material on the Internet.”); Shielding Children’s Retinas
from Egregious Exposure on the Net (SCREEN) Act, S.
5259, 117th Cong. (2022); Adult Nudity and Sexual Activity,
Meta,       https://transparency.fb.com/policies/community-
standards/adult-nudity-sexual-activity [https://perma.cc/
SJ63-LNEA] (“We restrict the display of nudity or sexual
activity because some people in our community may be
sensitive to this type of content.”). While platforms may or
may not share Congress’s moral concerns, they have
independent commercial reasons to suppress sexually
explicit content. “Such alignment does not transform private
conduct into state action.” O’Handley, 62 F.4th at 1157.
    CHD insists that it “is not arguing that Section 230 turns
all content moderation by all websites into state action,” but
rather that “Section 230(c)(2), in combination with . . .
sustained federal pressure” and “statements of strong
preference” and “encouragement,” turns Meta’s handling of
vaccine-related content into state action. As we have
explained, those statements and requests do not establish
either coercion or joint action. That Section 230 operates in
the background to immunize Meta if it chooses to suppress
vaccine misinformation—whether because it shares the
government’s health concerns or for independent
commercial reasons—does not transform Meta’s choice into
state action.
    If we were to accept CHD’s argument, it is difficult to
see why would-be purveyors of pornography would not be
able to assert a First Amendment challenge on the theory
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.        31


that, viewed in light of section 230, statements from
lawmakers urging internet providers to restrict sexually
explicit material have somehow made Meta a state actor
when it excludes constitutionally protected pornography
from Facebook. So far as we are aware, no court has ever
accepted such a theory.
                       *       *       *
     CHD’s inability to establish state action is fatal to all of
its First Amendment claims—for damages under Bivens, for
declaratory relief, and for an injunction. And to the extent
that CHD continues to argue on appeal that Meta’s disabling
of its donation button was a “taking” under the Fifth
Amendment, that claim fails for the same reason.
    Meta identifies several other hurdles to CHD’s damages
claims. For example, it argues that CHD cannot hold Meta,
a private corporation, liable under Bivens, see Correctional
Servs. Corp. v. Malesko, 

 (2001); that CHD
has not adequately alleged that Zuckerberg was personally
involved in any alleged constitutional violation, see Iqbal,

; and that it would be inappropriate to extend
Bivens to this context, see Egbert v. Boule, 

,
491–93 (2022); Pettibone v. Russell, 

, 454–55
(9th Cir. 2023). Because the state-action inquiry resolves all
of the constitutional causes of action, we need not reach
those issues.
     Our decision should not be taken as an endorsement of
Meta’s policies about what content to restrict on Facebook.
It is for the owners of social media platforms, not for us, to
decide what, if any, limits should apply to speech on those
platforms. That does not mean that such decisions are wholly
unchecked, only that the necessary checks come from
competition in the market—including, as we have seen, in
32      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


the market for corporate control. If competition is thought to
be inadequate, it may be a subject for antitrust litigation, or
perhaps for appropriate legislation or regulation. But it is not
up to the courts to supervise social media platforms through
the blunt instrument of taking First Amendment doctrines
developed for the government and applying them to private
companies. Whether the result is “good or bad policy,” that
limitation on the power of the courts is a “fundamental fact
of our political order,” and it dictates our decision today.
Lugar, 

.
                              III
    CHD claims that “the warning label and fact-checks”
Meta placed on its posts violated the Lanham Act. The
district court dismissed that claim because it held that
(1) CHD’s alleged injuries did not fall within the zone of
interests that the Lanham Act protects and (2) the fact-
checking labels were not statements made in “commercial
advertising or promotion.” We agree with the district court
on the latter ground, so we need not reach the former.
    As relevant here, the Lanham Act provides a cause of
action against any person who, “in commercial advertising
or promotion, misrepresents the nature, characteristics,
qualities, or geographic origin of his or her or another
person’s goods, services, or commercial activities.” 

(a)(1)(B). We have defined “commercial
advertising or promotion” to encompass “(1) commercial
speech, (2) by a defendant who is in commercial competition
with [the] plaintiff, (3) for the purpose of influencing
consumers to buy [the] defendant’s goods or
services, . . . (4) that is sufficiently disseminated to the
relevant purchasing public.” Ariix, LLC v. NutriSearch
Corp., 

, 1115 (9th Cir. 2021). “Commercial
       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.     33


speech,” we have explained, generally refers to speech that
“‘does no more than propose a commercial transaction.’” 

(quoting United States v. United Foods, Inc., 

 (2001)).
    By that definition, Meta did not engage in “commercial
speech”—and, thus, was not acting “in commercial
advertising or promotion”—when it labeled some of CHD’s
posts false or directed users to fact-checking websites.
Meta’s commentary on CHD’s posts did not represent an
effort to advertise or promote anything, and it did not
propose any commercial transaction, even indirectly.
    In arguing to the contrary, CHD emphasizes that we have
looked to the “economic motivation” of the speaker in
assessing whether speech is commercial in nature. Ariix, 985
F.3d at 1116 (quoting Hunt v. City of Los Angeles, 

 (9th Cir. 2011)). According to CHD, Meta placed
labels on its posts in order to “promot[e]” Meta’s fact-
checkers, who compete with CHD “in the nonprofit health
information market.” It also says that Meta sought to fact-
check CHD’s posts to ensure that it continued to receive
advertising revenue from vaccine manufacturers and to
dissuade lawmakers from repealing section 230—“which is
worth billions of dollars” to Meta. But economic motivation
is a factor we consider “[w]here the facts present a close
question,” which the facts here do not. Hunt, 

; see Dex Media W., Inc. v. City of Seattle, 

 (9th Cir. 2012) (explaining that an “economic motive in
itself is insufficient to characterize a publication as
commercial”). More importantly, the economic-motivation
test asks “whether the speaker acted primarily out of
economic motivation, not simply whether the speaker had
any economic motivation.” Ariix, 985 F.3d at 1116. As we
have explained, “[a] simple profit motive to sell copies of a
34      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


publication or to obtain an incidental economic benefit,
without more, does not make something commercial speech.
Otherwise, virtually any newspaper, magazine, or book for
sale could be considered a commercial publication.” Id. at
1117.
    Under any of CHD’s theories, the allegations suggest at
most that Meta acted with an economic motivation “to
obtain an incidental economic benefit.” Ariix, 985 F.3d at
1117. As described in the complaint, Meta’s economic
interests are far too remote from the challenged speech for it
to be plausible that “the economic benefit was the primary
purpose for speaking.” Id. The district court therefore
correctly concluded that the complaint did not state a claim
under the Lanham Act.
                              IV
    CHD also asserts a civil RICO claim. RICO makes it a
crime for a person employed by or associated with an
enterprise to conduct or participate in the conduct of the
enterprise’s affairs through a pattern of racketeering activity,
and it allows “[a]ny person injured in his business or
property by reason of a violation” to bring a civil damages
action. 

(c); see 

 § 1962(c). As relevant
here, the “racketeering activity” covered by RICO includes
“any act which is indictable under” the federal wire fraud
statute, 

. 

 § 1961(1)(B). That statute, in
turn, prohibits the use of electronic communications for the
purpose of executing “any scheme or artifice to defraud, or
for obtaining money or property by means of false or
fraudulent pretenses, representations, or promises.” Id.
§ 1343.
    To survive a motion to dismiss, CHD needed to plead
facts that would support a plausible inference that the
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.       35


defendants had engaged in a scheme or artifice to defraud
and that it suffered injury by reason of that scheme. It did not
do so. In the complaint, CHD described a scheme whereby
Meta placed warning labels on CHD’s posts with the intent
to “clear the field” of CHD’s alternative point of view, thus
keeping vaccine manufacturers in business so that they
would buy ads on Facebook and ensure that Zuckerberg
obtained a return on his investments in vaccine technology.
CHD has now abandoned that theory and instead focuses on
a different theory that it advanced for the first time in
response to the motion to dismiss. Under that theory, the
object of the scheme was “to deceive visitors to CHD’s
Facebook page into giving their charitable dollars not to
CHD, but to other, competing nonprofit organizations.” The
district court might have deemed that theory to be forfeited,
but because it addressed the theory on the merits, we will do
so as well.
    CHD emphasizes that a RICO plaintiff alleging fraud
need not show that it relied on any false statements by the
defendant but can in some cases allege that the defendant
harmed it by deceiving third parties. For example, in Bridge
v. Phoenix Bond & Indemnity Co., the Supreme Court held
that losing bidders in a tax-lien auction could bring a RICO
action against rival bidders who engaged in a fraudulent
scheme to win auctions by deceiving the seller. 

, 649–50 (2008). The Court offered an example to
illustrate the point: “[S]uppose an enterprise that wants to get
rid of rival businesses mails misrepresentations about them
to their customers and suppliers . . . . If the rival businesses
lose money as a result of the misrepresentations, it would
certainly seem that they were injured in their business ‘by
reason of’ a pattern of mail fraud.” 

 (quoting 

(c)).
36      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


    The rule in Bridge does not help CHD because the key
deficiency in CHD’s claims of wire fraud is not that the
alleged deception targeted third parties; it is the disconnect
between the alleged deception and the asserted injury to
CHD. The statutory phrase “by reason of” requires
proximate cause. Holmes v. Securities Inv. Prot. Corp., 

 (1992). For RICO purposes, that means the
plaintiff must allege “some direct relation between the injury
asserted and the injurious conduct alleged.” Hemi Grp., LLC
v. City of New York, 

 (2010) (quoting Holmes,

). For example, in Bridge, the bidders
adequately pleaded proximate cause because their auction
losses were the “direct result” of the alleged fraud. 

. The rival bidders deceived the seller about the share
of tax liens for which they were eligible, thereby reducing
the losing bidders’ share. 

 at 643–44, 658. “[N]o
independent factors” accounted for the plaintiffs’ loss. 

.
    More recently, in Hemi Group, LLC v. City of New York,
the Supreme Court held that a RICO plaintiff failed to plead
proximate cause where a direct link was lacking. 

. The City claimed that a cigarette vendor committed fraud
by neglecting to file required reports listing its purchasers
with the State, obstructing the City’s efforts to collect taxes
from the unidentified purchasers. 

 at 5–6. The Court
rejected the claim because the conduct “directly responsible”
for the City’s injury—the purchasers’ failure to pay taxes—
was “distinct from the conduct giving rise to the fraud”—the
vendor’s failure to file the purchaser reports. 

; see
Anza v. Ideal Steel Supply Corp., 

 (2006).
Unlike in Bridge, the losses to the City flowed from the
“independent actions” of purchasers to withhold the taxes
they owed. Hemi Grp., 

.
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.       37


    The causal chain that CHD proposes is, to put it mildly,
indirect. CHD contends that Meta deceived Facebook users
who visited CHD’s page by mislabeling its posts as false.
The labels that Meta placed on CHD’s posts included links
to fact-checkers’ websites. If a user followed a link, the fact-
checker’s website would display an explanation of the
alleged falsity in CHD’s post. On the side of the page, the
fact-checker had a donation button for the organization.
Meanwhile, Meta had disabled the donation button on
CHD’s Facebook page. If a user decided to donate to the
fact-checking organization, CHD maintains, that money
would come out of CHD’s pocket, because CHD and fact-
checkers allegedly compete for donations in the field of
health information.
    The alleged fraud— Meta’s mislabeling of CHD’s
posts— is several steps removed from the conduct directly
responsible for CHD’s asserted injury: users’ depriving
CHD of their donation dollars. At a minimum, the sequence
relies on users’ independent propensities to intend to donate
to CHD, click the link to a fact-checker’s site, and be moved
to reallocate funds to that organization. This causal chain is
far too attenuated to establish the direct relationship that
RICO requires. Proximate cause “is meant to prevent these
types of intricate, uncertain inquiries from overrunning
RICO litigation.” Anza, 

.
    CHD’s theory also strains credulity. It is not plausible
that someone contemplating donating to CHD would look at
CHD’s Facebook page, see the warning label placed there,
and decide instead to donate to . . . a fact-checking
organization. See Twombly, 

. The district
court noted that CHD did not allege that any visitors to its
page had in fact donated to other organizations because of
Meta’s fraudulent scheme. CHD is correct that an actual
38     CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


transfer of money or property is not an element of wire fraud,
as “[t]he wire fraud statute punishes the scheme, not its
success.” Pasquantino v. United States, 

(2005) (alteration in original) (quoting United States v.
Pierce, 

 (2d Cir. 2000)). But the fact that
no donations were diverted provides at least some reason to
think that no one would have expected or intended the
diversion of donations.
    If that were not enough, the Supreme Court has
cautioned us to ensure that fraud offenses be defined “with
sufficient definiteness that ordinary people can understand
what conduct is prohibited and . . . in a manner that does not
encourage arbitrary and discriminatory enforcement.”
Skilling v. United States, 

, 402–03 (2010)
(quoting Kolender v. Lawson, 

 (1983)); see
McDonnell v. United States, 

, 576 (2016)
(noting a due-process concern with the prospect of
“prosecution, without fair notice, for the most prosaic
interactions”). In seeking to hold the defendants liable for
statements on matters of public concern, CHD ignores that
caution. For example, under CHD’s view, it would seem that
a political party could bring a RICO claim against a rival
political party on the theory that its allegedly false
statements were part of a fraudulent scheme to divert
political contributions from the plaintiff party to its rival.
Such an application of the fraud statutes would raise serious
First Amendment concerns. We reject CHD’s invitation to
construe fraud so broadly.
                              V
   We affirm the district court’s dismissal of CHD’s claims
against Science Feedback for insufficient service of process.
Although the dismissal was without prejudice, we have
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.      39


jurisdiction over CHD’s appeal. Unlike a dismissal with
leave to amend, which permits further proceedings and
therefore is not final, see WMX Techs., Inc. v. Miller, 

, 1136 (9th Cir. 1997) (en banc), the dismissal here
means that the case “is over as far as the district court is
concerned,” so it is final and appealable, De Tie v. Orange
County, 

 (9th Cir. 1998); see Constien
v. United States, 

 (10th Cir. 2010)
(“[D]ismissal without prejudice for failure of service is a
dismissal of the action and not just the complaint because no
amendment of the complaint could cure the defect.”).
    We review the district court’s assessment of the
adequacy of service for abuse of discretion, and we find
none. See Rio Props., Inc. v. Rio Int’l Interlink, 

 (9th Cir. 2002). CHD made two efforts to serve
Science Feedback, but both were unsuccessful. It then asked
the district court to let it serve Meta’s counsel instead,
arguing that the contractual relationship between Meta and
Science Feedback made such service appropriate. The
district court denied that motion but stated that CHD could
renew it if further efforts to serve proved ineffective. CHD
never did so. Although it made another unsuccessful attempt
at service, it did nothing else until after the district court
entered judgment.
    CHD argues that Science Feedback has had actual notice
of this litigation, but that is not a substitute for service of
process under Federal Rule of Civil Procedure 4. See
Jackson v. Hayakawa, 

 (9th Cir. 1982).
It also contends that the district court should have applied
Rule 4(m), which requires that a court provide notice to the
plaintiff before dismissing the action on its own motion
when service has not been timely made. But Science
40       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


Feedback is domiciled in France, and by its terms, Rule 4(m)
“does not apply to service in a foreign country.”
    The motions for judicial notice (Dkt. Nos. 64, 70, 78, 86,
92) are DENIED.
     AFFIRMED.


COLLINS, Circuit Judge, concurring in part, concurring in
the judgment in part, and dissenting in part:

    I believe that Children’s Health Defense (“CHD”) has
shown that it could plausibly allege a First Amendment
claim for injunctive relief against Defendant Meta Platforms
Inc. (“Meta”), 1 and I therefore respectfully dissent from the
majority’s contrary conclusion. However, I agree that all of
CHD’s other claims were properly dismissed, and I therefore
concur in the judgment as to those remaining claims and in
Parts III, IV, and V of the majority opinion.
                                   I
                                   A
    Before sketching the facts that I take as true for purposes
of this appeal, I first address an important threshold question
concerning what factual allegations we may properly
consider.
   Because this appeal arises from a district court order
granting a motion to dismiss for failure to state a claim, we
must take the well-pleaded allegations of the operative

1
  Meta was known as “Facebook, Inc.” until October 2021. For
convenience, I will refer to it consistently as “Meta,” even with respect
to events before that date.
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.      41


complaint as true and draw all reasonable inferences in favor
of CHD. Shields v. Credit One Bank, N.A., 

,
1220 (9th Cir. 2022). We must “likewise take as true for
purposes of this appeal the additional well-pleaded
contentions” contained in any materials that were submitted
to the district court as reflecting the substance of a proposed
amendment to the complaint. Broidy Cap. Mgmt., LLC v.
State of Qatar, 

, 586 (9th Cir. 2020).
    However, CHD has also submitted certain additional
materials for the first time in this court, and the parties
sharply disagree as to whether, and to what extent, we may
consider these materials in assessing the legal sufficiency of
CHD’s claims. Under the unique circumstances of this case,
I agree with CHD that we should take judicial notice of the
existence of certain new, highly relevant documents that
have only recently become available and that, like the
materials submitted by CHD to the district court, effectively
reflect specific additional factual allegations that CHD
proposes to plead if it is given leave to amend on a remand.
     I recognize that, on appeal from the dismissal of a
complaint, a plaintiff generally cannot suggest new grounds
for amending the complaint for the first time in this court.
See, e.g., Riggs v. Prober & Raphael, 

(9th Cir. 2012). However, CHD does not purport to add
wholly new legal theories or claims, but rather only
additional factual allegations in support of its existing
claims. Moreover, its newly suggested amendments are
limited to factual allegations based on documents that were
concededly unavailable to CHD at the time of the district
court proceedings and that have only become subsequently
available through compulsory processes employed in other
litigation or in legislative investigations or through Freedom
of Information Act requests. We can take judicial notice of
42      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


the limited fact that these documents exist and have become
available to CHD during the course of this appeal. See Lee
v. City of Los Angeles, 

 (9th Cir. 2001).
While their contents cannot be judicially noticed for their
truth, see 

 CHD may properly draw on them in sketching
the additional factual allegations that it could now make if it
is given leave to amend.
    I disagree with Meta’s suggestion that this court’s only
procedurally proper option is to order a limited remand to
the district court so that that court can first consider the
wholly legal question of the viability of CHD’s claims in
light of these new potential allegations, after which we
would then review the matter de novo. Cf. FED. R. CIV. P.
62.1; FED. R. APP. P. 12.1. While we could certainly insist
that CHD proceed in that fashion, it is within our discretion,
under these unique circumstances, to simply consider the
legal sufficiency of such additional allegations ourselves in
evaluating whether CHD can state a claim on which relief
may be granted. The limited remand suggested by Meta here
would be a pointless waste of time and judicial resources and
would needlessly further postpone our obligation to decide
the novel, difficult, and important legal questions raised by
this appeal. In my view, given the weighty First Amendment
interests at stake in this case and the considerable difficulties
inherent in attempting to uncover facts concerning alleged
behind-the-scenes interactions between Meta and
Government personnel, we should exercise our discretion in
favor of considering the significance of the additional
allegations CHD could make in light of these newly
available documents.
       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.     43


                             B
    With these principles in mind, I take the following
factual contentions as true for purposes of this appeal from
the district court’s order dismissing CHD’s claims at the
pleadings stage.
                             1
    CHD is a Georgia-based non-profit membership
organization founded in 2015 by Robert F. Kennedy, Jr.
(“Kennedy”), who remains its chairman. Its professed
mission is “to educate the public about the risks and harmful
effects of chemical exposures upon prenatal and children’s
health, including from particular vaccines and
environmental health hazards, such as 5G and wireless
networks and products, and to advocate for social change
both legislatively and through judicial action.” “CHD’s
primary sources of revenue derive from membership dues
and donations that CHD solicits on its website and, formerly,
on its Facebook page.”
    With respect to vaccines, “CHD advocates for open and
honest public debate on the efficacy and safety of the . . .
entire Child and Adolescent Immunization Schedule” of the
Centers for Disease Control and Prevention (“CDC”). CHD
is sharply critical of the CDC’s vaccine policies, stating on
its website that “the CDC has become a mouthpiece for [the
pharmaceutical] industry and has protected the ‘all vaccines
for all children’ policy despite peer-reviewed science to the
contrary.” Indeed, CHD argues that the CDC has become so
plagued by conflicts of interest that the subject of “vaccine
safety should be taken from the CDC” altogether. CHD’s
website contains links to numerous articles concerning
vaccines and other topics, including both advocacy pieces
44     CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


and scientific studies from “peer-reviewed, published
journals.”
    Meta is a California-based corporation that operates,
among other things, two large social media platforms,
namely, Facebook and Instagram. According to the
operative complaint, Facebook has “214 million users in the
United States and 2.2 billion worldwide.” Mark Zuckerberg
is Meta’s co-founder and its chairman, CEO, and controlling
shareholder. As relevant here, Facebook enables users to
create webpages on which they can share information,
engage in advocacy, and solicit donations. Facebook users
can also choose to “follow” other users’ Facebook pages that
are of interest to them. In November 2017, CHD agreed to
Facebook’s terms of service and created its own Facebook
page. By late 2020, CHD’s Facebook page, which it used to
promote its views on vaccines and other matters, had more
than 122,000 followers.
                             2
     CHD alleges that, even before the Covid pandemic and
the development of Covid vaccines, CHD’s general
advocacy concerning vaccine safety drew the attention of
Government officials, who sought to pressure Meta to delete,
or at least to reduce the visibility of, what those officials
contended was “vaccine misinformation.” In particular,
CHD points to a February 14, 2019 letter from Congressman
Adam Schiff to Meta asking it to identify what measures it
currently took to address “misinformation related to
vaccines on [its] platforms” and “encourag[ing] [it] to
consider what additional steps [it] can take to address this
growing problem.” CHD alleges that, while ostensibly a
strictly informational inquiry, Congressman Schiff’s letter
must be understood against a larger backdrop in which
         CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.                45


various legislators, through hearings and public statements,
sought to pressure social media companies to delete or
restrict a variety of different categories of disfavored
content. CHD points, in particular, to April 2019 public
remarks by House Speaker Nancy Pelosi in which she raised
the possibility of removing the immunity for hosting third-
party content that is granted to social media platforms by
§ 230 of the Communications Act of 1934. 2 In those
remarks, the Speaker noted that, when the subject of § 230
is raised with social media companies, “you really get their
attention,” and she stated that it was “not out of the question”
that § 230’s immunity “could be removed” by Congress. As
she explained, “for the privilege of 230, there has to be a
bigger sense of responsibility” on the part of social media
companies.
     Three weeks after Congressman Schiff’s letter, Meta
announced it was taking a variety of steps to reduce the
visibility of “misinformation about vaccinations.” On April
26, 2019, Meta also announced that it would remove
“fundraising tools” from “Pages that spread misinformation
about vaccinations on Facebook.” In accordance with that
policy, Meta deactivated the fundraising function on CHD’s
Facebook page six days later. Around the same time, Meta
“permanently disabled the ‘dispute’ function on CHD’s
account so that neither CHD” nor Kennedy “could
challenge,” “through direct submission,” Meta’s actions
against CHD. Although CHD and Kennedy, of course, could
still send “written requests” to Meta objecting to its actions,
Meta consistently “ignored” these requests. Meta also took

2
  I discuss in detail below the nature of this immunity and the critical
practical role it plays in making possible the sorts of gigantic platforms
operated by Meta. See infra section III.
46     CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


steps to block the posting of some vaccine-related content,
including on CHD’s Facebook page. For example, on June
9, 2019, Meta blocked CHD from displaying, on its
Facebook page, a videotape of an interview in which
Kennedy “discuss[ed] a pending lawsuit against Merck &
Co.” concerning its Gardasil vaccine. On September 4,
2019, Meta also added warnings directly onto CHD’s
Facebook page, stating that “[t]his Page posts about
vaccines” and that those who want “reliable, up-to-date
information” about vaccines should “[g]o to CDC.gov,” the
webpage of the CDC.
                             3
    After the growing Covid pandemic resulted in
widespread lockdowns and societal disruption beginning in
March 2020, many public officials began to express a
focused concern over Covid-related “misinformation” on
social media. For example, in early June 2020, the House
Speaker sharply criticized social media platforms for failing
to halt the spread of “COVID-19 disinformation.” Later that
month, two subcommittees of the House Committee on
Energy and Commerce held a joint hearing on
“Disinformation Online.” In his opening remarks, the
chairman of one subcommittee stated that social media
platforms had “become awash in disinformation,” including
“lies about COVID 19.” He stated that the “status quo is
unacceptable,” and that, “[w]hile Section 230 has long
provided online companies the flexibility and liability
protections they need to innovate and to connect people from
around the world, it has become clear that reform is
necessary if we want to stem the tide of disinformation
rolling over our country.”        The chair of the other
subcommittee stated, in her opening remarks, that “Section
230” had come to effectively “protect[] business models that
       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.      47


generate profits off scams [and] fake news”; that this was
“never the intent” of Congress; that, “since both courts and
industry refuse to change, Congress must act”; and that she
“look[ed] forward to working with [her] colleagues to
modernize Section 230.”
    In September 2020, Zuckerberg stated in an interview
that Meta was actively working with the CDC and the World
Health Organization (“WHO”) “to remove clear
misinformation about health-related issues that could cause
an imminent risk of harm.” In October 2020, Zuckerberg
and the then-CEOs of Twitter, Inc. and Alphabet Inc. (which
operates the various Google products) were subpoenaed to
testify at an October 28 hearing of the Senate Commerce
Committee entitled, “Does Section 230’s Sweeping
Immunity Enable Big Tech Bad Behavior?”
     In early 2021, as Covid vaccines were first becoming
widely available, various Executive Branch officials took
steps to address a specific concern about what they
considered to be “misinformation” about these new Covid
vaccines, as well as about Covid more generally. These
officials included Robert Flaherty, Deputy Assistant to the
President and White House Director of Digital Strategy, and
Andrew Slavitt, who served as a White House Senior
Advisor for the COVID-19 Response. Shortly after joining
the new Administration, Flaherty reached out to Meta to
inquire about its policies concerning Covid-related
information on its platforms. On February 9, 2021, Meta
responded by email to Flaherty with its “responses to [his]
initial questions.” In response to Flaherty’s specific inquiry
as to how Meta handled Covid-related claims “that are
dubious, but not provably false,” Meta stated that, while its
practice was to “remove claims public health authorities tell
us have been debunked or are unsupported by evidence,” it
48     CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


also took measures to limit the distribution of content that
“contributes to unfounded hesitancy towards the COVID-19
vaccine,” even if such content “does not qualify for removal”
(emphasis added). Meta also stated that, where information
did warrant removal under its policies, “multiple violations”
would lead to restrictions on the relevant Facebook account,
including “suspend[ing] the entire Page, Group, or account.”
The email assured Flaherty that Meta “will begin enforcing
this policy immediately.” The next day, February 10, Meta
took down Kennedy’s Instagram account.
    In a March 21, 2021 email to Slavitt, Meta confirmed
that it would make a specific named employee “available on
a regular basis” to interface with the White House, noting
that the employee had already “been coordinating the
product work that matters most to your teams.” In that same
email, Meta confirmed that, in response to Slavitt’s prior
inquiry about the available “levers for reducing virality of
vaccine hesitancy content,” Meta would make “additional
changes that were approved late last week and that [it would]
be implementing over the coming weeks.” These included
“reducing the virality of content discouraging vaccines that
does not contain actionable misinformation” and removing
“Groups, Pages, and Accounts” that posted vaccine-related
content that, while truthful, was “sensationalized.”
    Four days later, two subcommittees of the House Energy
and Commerce Committee again held a joint hearing on
“disinformation” on social media platforms, and Zuckerberg
and the CEOs of Alphabet and Twitter all testified at this
hearing. In his opening remarks, one of the subcommittee
chairs stated that he was concerned about, among other
things, “antivaxxers, COVID deniers, QAnon supporters,
and Flat earthers.” Disinformation Nation: Social Media’s
Role in Promoting Extremism and Misinformation, Virtual
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.       49


Joint Hearing Before the Subcomm. on Commc’ns & Tech.
& the Subcomm. on Consumer Prot. & Com. of the H.
Comm. on Energy & Com., 117th Cong., SERIAL NO. 117-
19, at 2 (Mar. 25, 2021). The chairman of the full committee,
in his opening statement, stated that “it is now painfully clear
that neither the market nor public pressure will force these
social media companies to take the aggressive action they
need to take to eliminate disinformation and extremism from
their platforms” and that “therefore, it is time for Congress
and this committee to legislate and realign these companies’
incentives.” Id. at 12.
    On April 13, 2021, Meta emailed Flaherty and Courtney
Rowe, another White House official, to follow up
concerning various questions they had raised about Meta’s
treatment of posts that might promote vaccine hesitancy.
Attached to this email were “Vaccine Hesitancy Examples,”
including one specifically from CHD’s Facebook page. The
email defined “vaccine hesitancy” content as including, inter
alia, truthful content that “discuss[es] the choice to vaccinate
in terms of personal and civil liberties or concerns related to
mistrust in institutions or individuals.” Meta explained that
it “utilize[s] a spectrum of levers for this kind of content,”
including “reducing the posts’ distribution, not suggesting
the posts to users, [and] limiting their discoverability in
Search.”
    On May 6, 2021, Flaherty emailed Meta to complain
about the inadequacy of Meta’s efforts to demote truthful
vaccine-hesitant content. He specifically complained that
Meta’s vaccine hesitancy policy was not “stopping the
disinfo dozen”—a group of 12 individuals, including
Kennedy, who were identified as spreading Covid
“misinformation” online. On May 12, Flaherty followed up
and complained that, as compared with other social media
50     CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


platforms, he thought that Meta was doing an inadequate job
in downgrading “anti-vaccine” content. He elaborated:

       But “removing bad information from search”
       is one of the easy, low-bar things you guys do
       to make people like me think you’re taking
       action. If you’re not getting that right, it
       raises even more questions about the higher
       bar stuff. . . .
       Youtube, for their warts, has done pretty well
       at promoting authoritative info in search
       results while keeping the bad stuff off of
       those surfaces. Pinterest doesn’t even show
       you any results other than official
       information when you search for “vaccines.”
       I don’t know why you guys can’t figure this
       out.

    Meta’s interactions with the Government extended
beyond the White House. In particular, on June 1, 2021,
Meta emailed the CDC, explaining that it had established a
“misinfo claims portal” in which selected CDC personnel
who had been “whitelisted” for access to the portal could
submit requests to have particular posts taken down from
Facebook. The cover email explained that Meta wanted to
ensure that “everyone who had been whitelisted” had “all the
info they need to start submitting claims.” The email
included an attached file explaining, in a set of slides, how
the “Facebook Content Request System: Government
Reporting System” worked. An authorized CDC employee
would use the designated URL for accessing the system—
www.facebook.com/xtakedowns/login—and then enter his
or her credentials. Once the user was logged into the system,
       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.     51


he or she could select from a menu of pre-programmed
reasons for requesting removal of the offending content,
such      as     “Covid     Misinformation,”       “Vaccine
Discouragement,” and “Covid Vaccine Misinformation.”
After selecting from among these options, the user would be
directed to submit “the relevant violating URLs” that the
user wanted taken down, up to a maximum of 20 in a single
report. After submitting the request, the user would receive
a confirmation email with a reference number to allow for
tracking and follow-up.
    On July 15, 2021, Surgeon General Vivek Murthy and
White House Press Secretary Jen Psaki appeared at a joint
press briefing concerning the Government’s response to
Covid. Among other things, Surgeon General Murthy asked
social media companies “to consistently take action against
misinformation super-spreaders on their platforms.” In an
apparent specific reference to the so-called disinformation
dozen—which specifically includes Kennedy—Psaki
referred to “12 people who are producing 65 percent of anti-
vaccine misinformation on social media platforms.” In an
apparent reference to Kennedy—who had been banned from
Meta’s Instagram platform—Psaki noted that these 12
“remain[ed] active on Facebook, despite some even being
banned on other platforms” that “Facebook owns.”
    The next day, July 16, Meta employees had a call with
the Surgeon General’s office to discuss Meta’s actions
against “health misinformation.” During the call, Meta
specifically touted its earlier enforcement against Kennedy,
claiming that, after he was banned from Instagram, “[h]e
then stopped posting on [Facebook] about vaccines at all.”
    One week later, on July 23, Meta sent an email to various
persons in the Department of Health and Human Services
52      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


(“HHS”), following up on a meeting with them earlier that
day to “take stock after the past week.” In particular, the
email summarized “steps taken to further address the
‘disinfo dozen,’” including removing 17 additional “Pages,
Groups, and [Instagram] accounts tied to the disinfo dozen,”
with the result that “every member of the disinfo dozen . . .
had at least one such entity removed.” Meta reiterated that
it had heard HHS’s “call for us to do more,” and it said that
it would reach out “to schedule the deeper dive on how to
best measure Covid related content.” Meta underscored that
it and HHS had “a strong shared interest to work together”
and that it would “strive to do all [it] can to meet our shared
goals.”
     On August 6, 2021, Meta employees stated, in an internal
email, that Meta was moving forward with a number of
recommendations for dealing with posts with Covid-related
“misinfo” or posts that were “misinfo adjacent.” The first,
listed as “Option 1a”, was to remove “assets linked to
Groups / Pages / Profiles / Accounts that have been removed
for    COVID       misinfo     violations”    from     users’
recommendations. As an example, Meta stated “RFK Jr.’s
[Instagram] Account is removed, so his [Facebook] Page
will be non-recommendable.”             This option was
recommended as a “stop-gap measure specifically targeting
Disinfo Dozen assets.”
   On August 17, 2022, Meta ultimately removed CHD
from its Facebook and Instagram platforms entirely.
                              C
    CHD filed this action in August 2020. On December 4,
2020, CHD filed its second amended complaint, alleging
violations of the First and Fifth Amendments, along with
violations of the Racketeer Influenced and Corrupt
       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.      53


Organizations Act (“RICO”) and the Lanham Act. The
complaint named as Defendants Meta, Zuckerberg, and two
of Meta’s so-called “fact-checking” organizations, Science
Feedback and the Poynter Institute for Media Studies, Inc.
(“Poynter Institute”). Seeking declaratory, injunctive, and
monetary relief, CHD alleged, inter alia, that Defendants
were either working in concert with, or under compulsion
from, the Federal Government to suppress CHD’s speech on
Meta’s platforms and to prevent CHD from fundraising on
those platforms.
    On June 29, 2021, the district court dismissed the
operative complaint, and the court entered final judgment the
next day. The district court dismissed Science Feedback
without prejudice for lack of service, and it dismissed the
remaining Defendants with prejudice for failure to state a
claim upon which relief may be granted. See FED. R. CIV. P.
12(b)(6). With respect to CHD’s constitutional claims, the
district court held that CHD had failed to allege sufficient
facts to raise a plausible inference that Meta or Zuckerberg
had “worked in concert with the CDC to censor CHD’s
speech, retaliate against CHD, or otherwise violate CHD’s
constitutional rights,” nor had CHD “alleged facts showing
government coercion sufficient to deem [Meta] or
Zuckerberg a federal actor.” Absent state action, the district
court held, any constitutional claims necessarily failed. The
district court dismissed the remaining claims in the case on
a variety of other grounds.
    CHD timely appealed the district court’s dismissal of its
suit. We have jurisdiction under 

.
                             II
   I agree that CHD cannot assert a Bivens claim against
Defendants for monetary damages based on alleged
54      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


violations of its First Amendment rights. See Egbert v.
Boule, 

, 498–99 (2022) (generally declining to
extend a Bivens remedy “to alleged First Amendment
violations”). But as we have recognized, claims for
injunctive and declaratory relief, unlike claims for damages,
do not rely on the Bivens cause of action. See Ministerio
Roca Solida v. McKelvey, 

 (9th Cir.
2016) (“The only remedy available in a Bivens action is an
award for monetary damages from defendants in their
individual capacities.” (citation omitted)). Rather, “[t]he
ability to sue to enjoin unconstitutional actions by state and
federal officers is the creation of courts of equity, and
reflects a long history of judicial review of illegal executive
action, tracing back to England.” Armstrong v. Exceptional
Child Ctr., Inc., 

, 327 (2015). Indeed, in
Correctional Services Corp. v. Malesko, 

 (2001),
the Supreme Court declined to recognize a Bivens remedy
against “a private corporation operating a halfway house
under contract with the Bureau of Prisons,” but it then went
on to note that “suits in federal court for injunctive relief”
remained available, because “injunctive relief has long been
recognized as the proper means for preventing entities”—
presumably including private parties that qualify as state
actors—“from acting unconstitutionally.” 

; see
also Lima v. U.S. Dep’t of Educ., 

, 1127–28 &
n.6 (9th Cir. 2020) (rejecting, based on Malesko, a Bivens
damages claim against the private corporate defendant, but
then rejecting on the merits the plaintiff’s claim for equitable
relief against that defendant as an alleged state actor acting
unconstitutionally).
    Of the various grounds for dismissal given by the district
court, the only one that would support rejecting CHD’s claim
for injunctive and declaratory relief concerning alleged First
          CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.                    55


Amendment violations is the district court’s conclusion that
Meta and Zuckerberg were not state actors for purposes of
the First Amendment. 3 As the Supreme Court has squarely
held, the First Amendment’s “Free Speech Clause prohibits
only governmental abridgment of speech” and “does not
prohibit private abridgment of speech.” Manhattan Cmty.
Access Corp. v. Halleck, 

, 808 (2019). To
qualify as a constitutional violation, therefore, a particular
deprivation of a constitutional right must be “fairly
attributable to the State.” Lugar v. Edmondson Oil Co., 

 (1982); see also Lindke v. Freed, 

 (2024). The Court has articulated a number of
alternative formulas under which conduct may be fairly
attributable to the state, with the details of those various tests
reflecting the relevant features of some of the distinct
specific contexts in which the state-action question has often
arisen. See Lugar, 457 U.S. at 938–39 (noting, for example,
the “public function test,” the “state compulsion test,” the
“nexus test,” and the “joint action test” (citations and internal
quotation marks omitted)). The ultimate inquiry, however,
remains whether the challenged acts are “attributable to the
State” in the sense that they are “traceable to the State’s
power or authority.” Lindke, 

.



3
  The district court dismissed the claims against Poynter solely on the
ground that it (like Meta) was a private corporation that, under Malesko,
could not be sued in a Bivens action. The district court also recognized,
however, that CHD’s allegations against Poynter were very limited and
that most of the challenged conduct was allegedly committed by Meta.
Against this backdrop, I think it is fair to say that the logic of the district
court’s no-state-action ruling as to Meta and Zuckerberg would
necessarily extend to Poynter as well, even if the district court does not
itself appear explicitly to have made this point.
56      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


    Because—as the variety of alternative formulas
underscores—the state-action inquiry often depends upon
specific features of the context at issue, I think it is important
to begin by first setting out in some detail the unique legal
context that provides the backdrop for this case. Although
the majority deems the entirety of a massive platform such
as Facebook as being in all respects the equivalent to a big
newspaper, see Opin. at 22, the analogy is not entirely apt.
As I shall explain in Section III, Meta’s truly gargantuan
platforms simply could not exist in anything resembling their
current form without the legal immunity that the Federal
Government has afforded to internet platforms under § 230
of the Communications Act of 1934, as added by the
Communications Decency Act of 1996. Thereafter, in
Section IV, I will explain how that context-specific feature
factors into the overall question of whether, on the specific
facts at issue here, CHD can adequately plead the requisite
state action.
                               III
    Meta is in the business of transmitting, on a vast scale,
the publicly available speech of others, primarily through its
Facebook and Instagram platforms. As quickly became
apparent in the early days of the internet, operating any such
open platform for the speech of third parties presents very
substantial liability risks that, if the platform became large
enough, would be practically impossible to manage or to
effectively mitigate. Congress’s solution was § 230, which
we have construed to provide broad immunity to internet
platforms hosting third-party speech.
                               A
   In assigning liability for transmitting defamatory
communications, the common law generally distinguishes
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.      57


among different classes of persons based on their role in
creating, or their knowledge of, the contents of those
communications. The greatest level of responsibility applies
to the “composer or original publisher of a defamatory
statement, such as the author, printer or publishing house,”
because that person “usually knows or can find out whether
a statement in a work produced by him is defamatory or
capable of a defamatory import.” RESTATEMENT (SECOND)
OF TORTS § 581 cmt. c. A lesser degree of responsibility is
applied to those, such as newsstands, bookstores, and
libraries, who are in the business of distributing large
volumes of expressive content prepared exclusively by third
parties. Because such persons cannot be expected to screen
in advance the content of every book, periodical, and article
they distribute, the common law assigns them liability for
defamation only if “there are special circumstances that
should warn the dealer that a particular publication is
defamatory.” Id. cmt. d. But no liability for defamation is
assigned to a person or entity, such a “telephone company,”
that “merely makes available to another equipment or
facilities that he may use himself for general communication
purposes.” Id. cmt. b. These three categories of persons
have sometimes been respectively referred to as
“publishers,” “distributors,” and “conduits,” see, e.g., Austin
v. CrystalTech Web Hosting, 

 (Ariz. Ct.
App. 2005); Eugene Volokh, Treating Social Media Like
Common Carriers?, 1 J. OF FREE SPEECH L. 377, 455 (2021),
and I will use that same shorthand here.
     Meta and others operating social media platforms do not
fit neatly into this taxonomy. Although in one sense they
merely provide “equipment or facilities” that third parties
may use “for general communication purposes,” the
facilities at issue here voluntarily disseminate those
58      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


communications in many cases to the world at large.
Because Meta knows, or can readily know, the content of
those communications, and is under no legal obligation to
transmit them, it cannot be classified as a mere conduit. Cf.
RESTATEMENT (SECOND) OF TORTS § 581 cmt. f (stating that
a person or entity (such as a telegraph company) that
provides communication-specific assistance in transmitting
a particular statement whose contents are known or
accessible to it may be liable for defamation if “he knows or
has reason to know that the message is libelous”); see also
id. § 612 (providing, however, a limited privilege that further
limits liability for such transmitters).
    In many senses, Meta most resembles a distributor,
because it transmits a truly enormous volume of third-party
content that could not feasibly be reviewed in advance and
that it plays no role in creating. But, as this case illustrates,
Meta (like many other platform operators) also manages the
content on its websites in ways that arguably go beyond that
of a traditional distributor, such as a bookstore or newsstand,
and that begin to resemble the actions of a publisher. It is
perhaps therefore unsurprising that, even in the early days of
the internet, at least one court concluded that a company
operating a “computer bulletin board” could be classified as
a “publisher,” rather than a “distributor,” if it “h[o]ld[s] itself
out to the public and its members as controlling the content”
of that platform and “implement[s] this control through [an]
automatic software screening program” or through
“Guidelines” that it uses to remove content. Stratton
Oakmont, Inc. v. Prodigy Servs. Co., 

, at
*1, 4 (N.Y. Sup. Ct. May 24, 1995). By opting for the
“benefits” of this heightened degree of “editorial control,”
Stratton Oakmont held, the platform at issue there “ha[d]
opened it up to a greater liability” than a mere distributor.
         CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.                 59

. Such a rule, of course, would likely spell the end
of the internet as we know it: for a variety of reasons,
virtually no one operating (or using) a platform would want
there to be no controls over what content may be posted, but
under Stratton Oakmont, the use of such controls could result
in crushing publisher-level liability for all third-party
content on the platform.
                                    B
    Congress promptly acted to resolve this problem by
adding a new § 230 to the Communications Act of 1934,
which has been classified as § 230 of the unenacted title 47
of the United States Code. 4 Section 230 accomplishes that
goal by first establishing certain rules limiting internet
platforms’ liability for posting or removing third-party
content and then expressly preempting any contrary state or
local law. See 

(e)(3) (“No cause of action
may be brought and no liability may be imposed under any
State or local law that is inconsistent with this section.”).




4
  Although we have frequently referred to the statute as “Section 230 of
the Communications Decency Act,” Enigma Software Grp. USA, LLC v.
Malwarebytes, Inc., 

, 670 (9th Cir. 2023); see also Fair
Hous. Council of San Fernando Valley v. Roommates.Com, LLC, 

 (9th Cir. 2008) (en banc), that is a misnomer. Title V
of the Telecommunications Act of 1996 is captioned as the
“Communications Decency Act of 1996,” see 

, 

, 133 (1996), and § 509 of that title added § 230 to the
Communications Act of 1934, which has been classified to the unenacted

. See 

 § 509, 110 Stat. at 137. The statute can thus
properly be referred to either as § 230 of the Communications Act of
1934 or as § 230 of Title 47, but not as § 230 of the Communications
Decency Act.
60     CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


Section 230’s general rules for limiting liability are
contained in subsection (c), which provides as follows:

      (c) Protection for “Good Samaritan”
      Blocking and Screening of Offensive
      Material
          (1) Treatment of Publisher or Speaker
            No provider or user of an interactive
            computer service shall be treated as the
            publisher or speaker of any information
            provided by another information
            content provider.
          (2) Civil Liability
            No provider or user of an interactive
            computer service shall be held liable on
            account of—
              (A) any action voluntarily taken in
              good faith to restrict access to or
              availability of material that the
              provider or user considers to be
              obscene, lewd, lascivious, filthy,
              excessively violent, harassing, or
              otherwise objectionable, whether or
              not such material is constitutionally
              protected; or
              (B) any action taken to enable or
              make available to information content
              providers or others the technical
         CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.                    61


                   means to restrict access to material
                   described in paragraph [(A)]. 5

(c). 6
     Section 230(c)(1) squarely rejects Stratton Oakmont by
flatly providing that no “interactive computer service shall
be treated as the publisher or speaker” of third-party content
that it hosts or transmits. 

(c)(1); see also 

§ 230(f)(2) (broadly defining an “interactive computer
service” as “any information service, system, or access
software provider that provides or enables computer access
by multiple users to a computer server”). By its terms, this
categorical rule applies regardless of whether the platform
uses the sort of controls to screen and remove content that
were at issue in Stratton Oakmont. But to be sure that
platforms would have the ability, inter alia, to use “blocking
and filtering technologies that empower parents to restrict
their children’s access to objectionable or inappropriate
online material,” id. § 230(b)(4) (declaring the statute’s
“policy”), § 230 goes further and prohibits platforms from
being held liable “on account of . . . any action voluntarily
taken in good faith to restrict access to or availability of
material that the provider or user considers to be obscene,
lewd, lascivious, filthy, excessively violent, harassing, or



5
  The statute actually says “paragraph (1),” but that is obviously a
scrivener’s error. See U.S. Nat. Bank of Oregon v. Independent Ins.
Agents of Am., Inc., 

 (1993).
6
  In its current form, the statute carves out certain specified exceptions
from its limitation on civil liability, such as for conduct that violates “any
law pertaining to intellectual property” or that violates certain sex-
trafficking laws. See 

(e)(2), (5)(A).
62      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


otherwise objectionable, whether or not such material is
constitutionally protected.” 

 § 230(c)(2)(A).
    This court has construed the resulting immunity
conferred by § 230 very broadly. We have held that
subsection (c)(1)’s rule that a platform operator shall not be
“treated as the publisher or speaker of any information
provided” by a third party does not merely prohibit the sort
of publisher-liability that was at issue in Stratton Oakmont.
See Barnes v. Yahoo!, Inc., 

 (9th Cir.
2009) (expressly rejecting the argument that, “because
Congress enacted section 230 to overrule Stratton Oakmont,
which held an internet service provider liable as a primary
publisher, not a distributor, the statute does no more than
overrule that decision’s application of publisher liability”
and that § 230 therefore leaves distributor liability intact).
Rather, we have held, § 230(c)(1) broadly “precludes courts
from treating internet service providers as publishers not just
for the purposes of defamation law, with its particular
distinction between primary and secondary publishers, but
in general.” Id. As we explained:

       [W]hat matters is not the name of the cause
       of action—defamation versus negligence
       versus intentional infliction of emotional
       distress—what matters is whether the cause
       of action inherently requires the court to treat
       the defendant as the “publisher or speaker” of
       content provided by another. To put it
       another way, courts must ask whether the
       duty that the plaintiff alleges the defendant
       violated derives from the defendant’s status
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.       63


       or conduct as a “publisher or speaker.” If it
       does, section 230(c)(1) precludes liability.

Id. at 1101–02.
     Under this reading of § 230, we held that a plaintiff’s
cause of action impermissibly treats an internet service
provider as the “publisher or speaker” of third-party content
if it necessarily rests on a duty concerning “reviewing,
editing, [or] deciding whether to publish or to withdraw from
publication third-party content.” Barnes, 

.
As noted above, the paradigmatic example of “a cause of
action that treats a website proprietor as a publisher” within
the meaning of § 230 “is a defamation action founded on the
[proprietor’s] hosting of defamatory third-party content,”
Doe v. Internet Brands, Inc., 

 (9th Cir.
2016), including hosting undertaken as a traditional
“publisher” or as a “distributor,” Barnes, 

.
But under the analysis we adopted in Barnes, § 230(c)(1)’s
immunity also extends to any duty that “would necessarily
require an internet company to monitor third-party content,”
HomeAway.com, Inc. v. City of Santa Monica, 

 (9th Cir. 2019) (emphasis added), or to remove such
content, Barnes, 

. As we stated in Barnes,
“removing content is something publishers do, and to
impose liability on the basis of such conduct necessarily
involves treating the liable party as a publisher of the content
it failed to remove.” 

 “Subsection (c)(1), by itself, shields
from liability all publication decisions, whether to edit, to
remove, or to post, with respect to content generated entirely
by third parties.” 

 (emphasis added).
   Barnes rejected the argument that, by construing
subsection (c)(1)’s immunity as extending to actions
concerning the monitoring and removal of content, we had
64     CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


rendered superfluous the distinct immunity set forth in
subsection (c)(2), which directly concerns potential liability
for “restrict[ing] access to or availability of material.” 

(c)(2); see Barnes, 

. We
explained that, unlike the immunity granted by § 230(c)(1),
the further immunity conferred by § 230(c)(2) would apply
to content that was partly created by the internet provider
itself and to access restrictions that went beyond what could
fairly be characterized as “publishing or speaking.” Barnes,

.
    Our broad construction of § 230 has been subject to
substantial criticism in a number of specific respects, but I
am bound by our settled precedent, and I do not question it
here. Moreover, regardless of any criticisms about the
precise scope of that immunity, the central point, for present
purposes, remains indisputable: § 230 confers a statutory
immunity without which Meta could not practicably operate
gigantic platforms such as Facebook and Instagram. The
potential liability for defamatory content alone—not to
mention other theories of platform host liability—would be
so crushing as to preclude the operation of these platforms
in anything resembling their current form. And, importantly,
the immunity granted by § 230 is purely an act of
congressional grace, because Meta has no plausible claim to
a constitutional entitlement to full immunity for publishing
or distributing constitutionally unprotected defamatory
content. Cf. Woodhull Freedom Found. v. United States, 

, 1299 (D.C. Cir. 2023) (holding that Congress’s
denial of § 230 immunity to internet-provider conduct that
amounts to aiding and abetting sex trafficking is not
overbroad or facially unconstitutional).
    In this respect, Meta’s position stands in sharp contrast
to that of a traditional publisher, such as a newspaper. A
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.      65


newspaper publisher’s editorial decisions over the third-
party content published in that paper are not broadly
immunized by statute from constitutionally permissible
liability. And because newspaper editors must, as a
consequence, consider the potential liability associated with
each third-party piece they publish, they necessarily limit
and individually select the third-party speech that they are
willing to include. In the absence of § 230’s immunity, Meta
would have to take comparable steps to manage and limit the
enormous potential liability that could arise from its
platforms’ hosting of third-party speech by behaving more
like a traditional newspaper. (At the very least, it would have
to behave more like a traditional newsstand or bookstore, if
one assumes, contrary to Stratton Oakmont, that the use of
algorithmic tools and of other content-management
measures is consistent with being a mere distributor rather
than a publisher.) But in all events, in a world without § 230,
Meta would almost certainly have to substantially reduce the
massive scale of its third-party content hosting; it would
presumably be much more pro-active than it already is about
screening out content; and it would be much more selective
about who it lets use its platforms and under what conditions.
But with § 230’s singular and broad immunity in place, Meta
is freed up to exercise practical—and potentially arbitrary—
control over the hosted content of the speech of more than
100 million people in the United States alone. In effect, by
virtue of the special treatment afforded under § 230 to its
massive platforms, Meta has been given the immunity of a
conduit for the billions of postings that (in conduit-like
fashion) it hosts, but that conduit-type immunity is coupled
with what, in many respects, is functionally the editorial
power of a publisher over everything on the platform.
66       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


     The truly gigantic scale of Meta’s platforms, and the
enormous power that Meta thereby exercises over the speech
of others, are thus direct consequences of, and critically
dependent upon, the distinctive immunity reflected in § 230.
That is, because such massive third-party-speech platforms
could not operate on such a scale in the absence of something
like § 230, the very ability of Meta to exercise such
unrestrained power to censor the speech of so many tens of
millions of other people exists only by virtue of the
legislative grace reflected in § 230’s broad immunity.
Moreover, as the above discussion makes clear, it was
Congress’s declared purpose, in conferring such immunity,
to allow platform operators to exercise this sort of wide
discretion about what speech to allow and what to remove.
In this respect, the immunity granted by § 230 differs
critically from other government-enabled benefits, such as
the limited liability associated with the corporate form. The
generic benefits of incorporation are available to all for
nearly every kind of substantive endeavor, and the limitation
of liability associated with incorporation thus constitutes a
form of generally applicable non-speech regulation. In sharp
contrast, both in its purpose and in its effect, § 230’s
immunity is entirely a speech-related benefit—it is, by its
very design, an immunity created precisely to give its
beneficiaries the practical ability to censor the speech of
large numbers of other persons. 7 Against this backdrop,

7
  The majority suggests that if § 230 were “enough for state action, every
large government contractor would be a state actor.” Opin. at 27.
However, as I shall explain below, I do not contend that § 230 alone
suffices to establish state action here. See infra at 85. But the majority
is also wrong in suggesting that the Government-conferred benefit here
is comparable to the others that it cites. Companies that are dependent
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.           67


whenever Meta selectively censors the speech of third
parties on its massive platforms, it is quite literally
exercising a government-conferred special power over the
speech of millions of others. The same simply cannot be
said of newspapers making decisions about what stories to
run or bookstores choosing what books to carry.
    I do not suggest that there is anything inappropriate in
Meta’s having taken advantage of § 230’s immunity in
building its mega-platforms. On the contrary, the fact that it
and other companies have built such widely accessible
platforms has created unprecedented practical opportunities
for ordinary individuals to share their ideas with the world at
large. That is, in a sense, exactly what § 230 aimed to
accomplish, and in that particular respect the statute has been
a success. But it is important to keep in mind that the vast
practical power that Meta exercises over the speech of
millions of others ultimately rests on a government-granted
privilege to which Meta is not constitutionally entitled.
                                IV
    In my view, this key fact—viz., that Meta is effectively
exercising a distinctive government-conferred power over
others’ speech when it decides whether and how to censor
third-party speech on its vast platforms—makes a crucial
difference in the state-action analysis. As I shall explain, the
particular state-action test that is most relevant here is the
one applied in Skinner v. Railway Labor Executives Ass’n,

 (1989). As relevant here, Skinner establishes
that, where a private party exercises a distinctive


on a “favorable regulatory environment” or on significant Government
contracts do not rely on a speech-related benefit that was purposely
created to facilitate the suppression of third parties’ speech.
68      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


government-conferred immunized power that is specifically
targeted at particular rights of third parties, and those
particular rights are ones that are protected from
governmental infringement by the Constitution, then that
private party’s interactions with the Government as to how
to exercise that power over those third parties’ constitutional
rights implicate constitutional standards and must comply
with those standards. And under that analysis, CHD can
adequately plead state action here.
                              A
     Skinner involved two sets of regulations that were
adopted to address the serious safety concerns presented by
intoxicated railroad workers. 489 U.S. at 608–09. The first
set, contained in “Subpart C” of the applicable regulations,
imposed mandatory drug testing on employees involved in
specified types of train accidents. 

. The second
set, in “Subpart D,” created a “permissive” regime of drug
testing that was available against persons who were not
covered by the mandatory provisions of Subpart C. 

 (emphasis added). Specifically, Subpart D authorized
railroads to require drug testing of (1) an employee as to
whom one or two supervisors had a “reasonable suspicion”
that the employee was under the influence; or (2) an
employee as to whom a supervisor had a reasonable
suspicion that the employee contributed to an accident’s
“occurrence or severity.” 

 Various organizations
representing railroad workers challenged these regulations
as a violation of the Fourth Amendment. 

. The
district court rejected these challenges, but we reversed. 

We concluded that, with the exception of the portion of
Subpart D that authorized drug tests upon reasonable
suspicion of impairment, the regulations did not require the
“showing of individualized suspicion” that we held was
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.      69


“essential” to conducting such a search under the Fourth
Amendment. 

. The Supreme Court then reversed
our judgment to the extent that it had invalidated Subpart D.
     At the outset, the Court had to address the threshold
question whether drug tests conducted under these
regulations implicated the protections of the Fourth
Amendment.         As the Court explained, “the Fourth
Amendment does not apply to a search or seizure, even an
arbitrary one, effected by a private party on his own
initiative,” but “the Amendment protects against such
intrusions if the private party acted as an instrument or agent
of the Government.” Skinner, 

. In applying
this overall standard, the Court first noted that the answer
was easy as to the mandatory testing requirements in Subpart
C: “A railroad that complies with the provisions of Subpart
C of the regulations does so by compulsion of sovereign
authority, and the lawfulness of its acts is controlled by the
Fourth Amendment.” 

 By contrast, the state-action issue
with respect to Subpart D required a more extensive analysis.
    As an initial matter, the Court explicitly rejected the
defendants’ argument that “the Fourth Amendment is not
implicated by Subpart D of the regulations, as nothing in
Subpart D compels any testing by private railroads.” 

 (emphasis added). As the Court explained, “[t]he
fact that the Government has not compelled a private party
to perform a search does not, by itself, establish that the
search is a private one.” 

. Even in the absence of
compulsion, a private party might be “deemed an agent or
instrument of the Government for Fourth Amendment
purposes” if the “degree of the Government’s participation
in the private party’s activities” was sufficient, “in light of
all the circumstances,” to trigger the Constitution’s
protections. 

 at 614–15.
70      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


     In concluding that “tests conducted by private railroads
in reliance on Subpart D” should not be viewed as being
“primarily the result of private initiative,” the Court
emphasized several considerations. First, the regulations in
Subpart D broadly preempted “state laws, rules, or
regulations covering the same subject matter,” including
rights recognized in collective bargaining agreements. 

. Indeed, the regulations specifically stated that
railroads could “not bargain away the authority to perform
tests granted by Subpart D.” 

 By these measures, “[t]he
Government ha[d] removed all legal barriers to the testing
authorized by Subpart D.” 

 Second, the regulations gave
the Government “the right to receive certain biological
samples and test results procured by railroads pursuant to
Subpart D.” 

 The Government had thereby “made plain
not only its strong preference for testing, but also its desire
to share the fruits of such intrusions.” 

 Third, the regime
created by Subpart D was compulsive vis-à-vis the
employee: “a covered employee” was not “free to decline his
employer’s request to submit to breath or urine tests under
the conditions set forth in Subpart D.” 

 These three
considerations—the Government’s conferral of a special
private power against others that was broadly immunized;
the Government’s interest in, and benefit from, the exercise
of that power; and the compulsive nature of that power when
wielded against other private parties—led the Court to
conclude that a railroad’s invocation of that power against
its employees was sufficiently done with “the Government’s
encouragement, endorsement, and participation” to
“implicate the Fourth Amendment.” 

 at 615–16.
    Having found that the Fourth Amendment applied to
searches conducted under Subpart D, the Court then
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.      71


concluded that the regulations did not violate the Fourth
Amendment. 

 at 633–34.
                              B
    Consideration of the same key three factors discussed in
Skinner strongly supports the view that Meta’s alleged
interactions with the Government here are sufficient to
implicate the First Amendment rights of CHD and those it
represents, including Kennedy.
    As I have explained, Meta here was not simply
exercising the normal editorial control that goes with being
an ordinary publisher who sifts through pre-publication
submissions and affirmatively decides what to include in its
publication. There are many such websites, and their
exercise of such conventional editorial judgment and
responsibility means that, from a practical point of view,
their very ability to exist and to operate does not depend
upon § 230’s grace (even if they are a beneficiary of it). By
contrast, Meta’s construction of massive and widely
available platforms for the hosting of the speech of
enormous numbers of third parties necessarily means that
those platforms exist and operate only by virtue of the
immunity conferred by § 230. Thus, the authority to manage
content on such mega-platforms is, in a very real sense, a
government-conferred power, and the Government, through
its broad preemption of “state laws, rules, or regulations
covering the same subject matter,” has intentionally
“removed all legal barriers” to Meta’s exercise of that power
over the speech of others. Skinner, 

. And, as
in Skinner, that government-conferred power is one that, by
its very design, is specifically directed at third-party rights
that are protected under the Constitution from encroachment
by the Government. In that sense, the immunized power
72       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


conferred is not akin, for example, to the generic benefits of
the liability limitations of the corporate form. Section 230,
by its structure and design, grants an immunized power
specifically directed at censoring the speech of others. 8
    Moreover, Meta’s exercise of that power is clearly
coercive from the point of view of the third parties whose
speech is targeted. Like the railway employees in Skinner,
they are not “free to decline” to have their speech removed
from the platform if Meta chooses to do so. Skinner, 

.
    The central question, then, is whether Skinner’s last
remaining factor—namely, governmental interest in, and
direct benefit from, specific exercises of that power—is
satisfied here. In addressing this factor, I think it is
important to note a critical difference between this case and
Skinner. In Skinner, the Government’s interest in, and
benefit from, the testing power conferred in Subpart D was
built into the regulations themselves, because those
regulations expressly gave the Government the right to
obtain certain results of those tests. 

 The same cannot be
said of the regime created by § 230, which provides for no
formal governmental role in the exercise of the powers that
it makes possible. Accordingly, unlike in Skinner, this
important state-action factor is not automatically satisfied
simply by virtue of the structure of the legal regime that the
Government has created. On that basis, the district court

8
  Meta is therefore wrong in suggesting that this case does not involve
the “exercise of some right or privilege created by the State.” O’Handley
v. Weber, 

, 1156 (9th Cir. 2023) (stating that a threshold
question in the state-action inquiry is “whether the alleged constitutional
violation was caused by the ‘exercise of some right or privilege created
by the State or by a rule of conduct imposed by the State or by a person
for whom the State is responsible’” (citing Lugar, 

)).
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.       73


below distinguished Skinner and held that it did not support
a finding of state action here. The majority relies on similar
reasoning, noting that § 230 neutrally protects whatever
editorial decisions Meta makes with respect to third-party
speech on its platforms. See Opin. at 29. But this reasoning
overlooks the possibility that, even though a governmental
benefit is not directly built into § 230’s legal regime, the
same relevant sort of governmental interest and benefit may
be supplied with respect to particular communications and
speakers by virtue of specific interactions between Meta and
the Government concerning such communications or
speakers. In view of the factual contentions summarized
earlier, that line has plainly been crossed here. In particular,
three distinct types of specific alleged interactions between
Meta and the Government, taken together, strongly confirm
the Government’s interest in, and benefit from, many of the
particular challenged exercises of that power. Skinner, 

.
    First, the above-described allegations confirm that high-
level Government officials made targeted requests, both
publicly and privately, for Meta to take action specifically
against the speech of CHD and Kennedy. In a private email,
Flaherty pointedly complained that Meta was not doing
enough to “stop[] the disinfo dozen,” which was a clear
reference to CHD and Kennedy. Psaki and Murthy likewise
publicly called for Meta and other platforms to target the
same “12 people who are producing 65 percent of anti-
vaccine misinformation.” In its private reassurances to
White House and other Executive Branch officials, Meta
repeatedly and specifically touted the targeted actions it had
taken against CHD and Kennedy. For example, in an email
to Flaherty, Meta attached a CHD Facebook post as an
example of the sort of truthful “vaccine hesitan[t]” speech
74      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


that it was targeting, just as it knew Flaherty wanted it to do.
The day after Murthy’s and Psaki’s press conference, Meta
followed up with Murthy’s office and, during that
conversation, specifically noted its targeted actions against
Kennedy’s vaccine-related speech. The following week, it
again emphasized, in discussions with HHS, the additional
steps it was taking against “the disinfo dozen.” On this
record, the Government expressed its specific interest in
suppressing particular speech of particular speakers—
including CHD and Kennedy—and Meta responded by
underscoring the steps it had taken, and planned to take, to
accomplish just that.
    Second, under the allegations here, Meta worked
extensively with Executive Branch officials to adjust and
refine its criteria and practices with respect to limiting or
suppressing vaccine-related speech. These were not simply
informational exchanges in which Meta passed along its
internal criteria for addressing such speech. Rather, Meta
engaged in a dialogue with Executive Branch officials to
develop and “begin enforcing” new policies with respect to
Covid-vaccine-related speech. In particular, there was
extensive discussion with Government officials about what
“levers” to exercise against truthful “vaccine hesitancy
content.” And the Government was hardly a passive
participant in these discussions. On the contrary, Flaherty
and others repeatedly chastised Meta for not doing enough
to suppress anti-vaccine content, unfavorably comparing
Meta to other social media companies and underscoring the
importance of Meta “mak[ing] people like me think you’re
taking action.” The allegations here raise a plausible
inference that Meta responded to such jawboning with
appeasing efforts at modifying its policies and practices with
respect to such speech.
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.       75


    Third, Meta went so far as to create an actual portal in
which pre-selected Government officials could log in and
then submit targeted requests for specific Covid-vaccine-
related posts to be taken down. On its face, this system
extended to truthful speech that the “whitelisted”
Government officials nonetheless deemed to promote
“Vaccine Discouragement.”
     It is also important to note that all of these actions took
place against a backdrop of continuous legislative threats, at
multiple levels, to limit or abolish the § 230 immunity upon
which Meta’s very ability to operate its mega-platforms
critically depends. These included congressional hearings in
both houses, at which Zuckerberg and other social media
CEOs were called to testify, as well as statements from high-
ranking officials including the House Speaker and relevant
committee chairs in both houses.                 Although, by
constitutional design, the legislative process is cumbersome,
and legislative threats are therefore harder to carry out than
others, the Speaker trenchantly observed that, when
legislators raise the subject of § 230 reform with social
media companies, “you really get their attention.” While I
agree that these various legislative comments and actions,
taken in isolation, do not themselves constitute
governmental compulsion of action under the traditional
“‘state compulsion’ test,” Lugar, 

; cf.
Kennedy v. Warren, 

, 1207–12 (9th Cir. 2023),
that is not dispositive. The Supreme Court held that state
action was present in Skinner even though compulsion was
concededly absent in that case. See Skinner, 

(finding state action even while agreeing that the
governmental regulations in Subpart D did not “compel[] a
private party to perform a search”). And these frequent and
high-level threats are certainly relevant in considering
76       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


whether, “in light of all the circumstances,” Meta’s
challenged actions here “are attributable to the Government
or its agents” under Skinner’s standards. 

 (citation
omitted).
    Taking these considerations together, the Government
“made plain” its “strong preference” for particular exercises
of Meta’s § 230-immunized power over third-party speech
on its mega-platforms. Skinner, 

. The
Government directly communicated to Meta its specific
interest in Meta acting to limit or remove (1) content that
expressed particular Government-disfavored viewpoints on
a specific subject (viz., vaccines in general and the Covid
vaccines in particular), and (2) the speech of CHD and
Kennedy on that subject in particular. With awareness of
that focused interest, and of the benefits that the Government
hoped to obtain if such speech were suppressed, Meta then
affirmatively worked with the Government to refine Meta’s
policies and practices concerning such speech in a way that
would be satisfactory to the Government, and it repeatedly
touted to the Government its specific actions directly
targeted against CHD and Kennedy. On these situation-
specific facts, I think that Skinner’s last remaining factor—a
governmental interest in, and benefit from, particular
exercises of the immunized power—is satisfied here. 9
    Accordingly, I would hold that, considered “in light of
all the circumstances,” Meta’s interactions with the
Government with respect to the suppression of specific
categories of vaccine-related speech, and in particular the

9
  A different and much more difficult state-action question would be
presented if Meta had refrained from such affirmative interactions with
the Government and instead was merely the passive recipient of criticism
or haranguing from Government officials.
         CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.                  77


speech of CHD and Kennedy, “suffice to implicate the
[First] Amendment.” Skinner, 

 (citation
omitted). Moreover, that conclusion makes perfect sense
when viewed from the converse perspective of what the
Government must not do when it interacts with such mega-
platforms. Having specifically and purposefully created an
immunized power for mega-platform operators to freely
censor the speech of millions of persons on those platforms,
the Government is perhaps unsurprisingly tempted to then
try to influence particular uses of such dangerous levers
against protected speech expressing viewpoints the
Government does not like. The Skinner-based analysis set
forth above properly recognizes that, when the Government
does so, and the platform operator responds
accommodatingly, the First Amendment is implicated.
Whether First Amendment standards have been violated was
not reached by the district court and therefore is not squarely
before us. 10
    Because I think that CHD could amend its complaint in
a manner that states a cause of action for injunctive and
declaratory relief based on the theory that Meta’s above-
described interactions with the Government implicate the
First Amendment rights of CHD, Kennedy, and CHD’s other
members, I would reverse the district court’s judgment in
favor of Meta to the extent it held to the contrary. 11 Because,

10
   As I note below, however, CHD’s allegations raise a plausible
inference that the Government sought to restrict CHD’s protected speech
for the illegitimate purpose of suppressing disfavored speech that
interfered with its policy objectives. See infra at 87.
11
  For many of the same reasons discussed above, CHD is clearly able to
plead sufficient facts to assert Article III standing to seek injunctive and
78       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


however, CHD’s showing on this score extends only to Meta
and not to Zuckerberg personally or to the Poynter Institute,
I would affirm the dismissal of the direct injunctive claims


declaratory relief against Meta. See Lujan v. Defenders of Wildlife, 

, 560–61 (1992) (noting that, at the pleading stage, the plaintiff
may rely on “mere allegations” to establish the core elements of
standing, which are (1) an injury in fact (2) that is fairly traceable to the
defendant’s challenged conduct and (3) that would be redressed by the
requested relief); cf. Murthy v. Missouri, 

, 1986 (2024)
(holding that, at the preliminary injunction stage, where “the parties have
taken discovery,” the plaintiff “must instead point to factual evidence”).
CHD has properly rested its standing both on injuries to itself and
injuries to members that it represents (such as Kennedy). See Hunt v.
Washington State Apple Advert. Comm’n, 

 (1977). In
contrast to Murthy, CHD has identified specific alleged instances in
which Government interaction with Meta led to “discrete instance[s]” of
censorship of CHD’s and Kennedy’s content. See Murthy, 144 S. Ct. at
1987. In February 2021, Meta responded to Slavitt’s inquiry about
restrictions on vaccine-hesitant content by stating that it would “begin
enforcing” a new “policy” on that score, and it then proceeded to take
down Kennedy’s Instagram account the very next day. Two months
later, Meta expressly reassured White House officials about the steps it
was taking against vaccine-hesitant content, and it specifically attached,
as an example, a post from CHD’s Facebook page. A month later, a
White House official complained that Meta was still not doing enough
to stop the vaccine-hesitant speech of the “disinfo dozen,” which
included Kennedy. Murthy and Psaki then singled out the same dozen
speakers, and Kennedy in particular, in their July 2021 press conference.
That was followed by Meta informing HHS officials, a week later, that
it had taken specific action against each one of the “disinfo dozen,”
including Kennedy, and thereafter Meta continued evaluating additional
restrictions on Kennedy. And because, in contrast to Murthy, CHD seeks
to enjoin the platform operator directly, it has “satisf[ied] traceability”
by alleging that Meta continues to exclude CHD’s and Kennedy’s posts
“under a policy that it adopted at the White House’s behest,” and an
injunction directed at Meta will redress that injury. See Murthy, 144
S. Ct. 1996–97 & n.11.
         CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.                  79


against those two defendants. 12 Of course, to the extent that
CHD were to establish an ultimate entitlement to injunctive
relief, Zuckerberg, the Poynter Institute, and others working
in concert with Meta might nonetheless be incidentally
covered by an injunction against Meta. And I would affirm,
under Egbert, CHD’s First-Amendment-based Bivens claim
for monetary damages.
                                    C
    None of the additional contentions raised by the majority
or by Defendants supports a contrary view with respect to
the state-action issue.
    Defendants rely heavily on our decision in O’Handley v.
Weber, 

 (9th Cir. 2023), in which we held that
state action was not present when a state election official
submitted a request to Twitter to remove a particular post
questioning the integrity of California’s elections. 

, 1160–61. But O’Handley was not presented with, and
did not consider, the points addressed here about the
significance of § 230 immunity under Skinner. Indeed,
O’Handley never even cited either § 230 or Skinner. As
such, O’Handley is distinguishable and not controlling here.
See Cooper Indus., Inc. v. Aviall Servs., Inc., 

 (2004) (“Questions which merely lurk in the record,
neither brought to the attention of the court nor ruled upon,
are not to be considered as having been so decided as to
constitute precedents.” (citation omitted)); Guerrero v. RJM
Acquisitions LLC, 

 (9th Cir. 2007)
(holding that “unstated assumptions on non-litigated issues

12
  I would likewise affirm the dismissal of CHD’s claim under the
Takings Clause. CHD has made no comparable showing of state action
with respect to its assertion that the disabling of its donate button on its
Facebook page was somehow a violation of the Takings Clause.
80       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


are not precedential holdings binding future decisions”
(citation omitted)).
    The majority concludes that the overall facts alleged here
do not plausibly reflect the sort of compulsion that the
caselaw typically requires to establish state action under the
state compulsion test. See Opin. at 22–27; cf. Bantam Books,
Inc. v. Sullivan, 

 (1963) (holding that actions
of Rhode Island state commission, which “exhort[ed]
booksellers” not to carry disfavored non-obscene titles,
violated the First Amendment where the commission’s
communications were “phrased virtually as orders,” were
“invariably followed up by police visitations,” and led
distributor to acquiesce in a manner that the lower courts
found “was not voluntary”). I am not entirely sure whether
the majority is correct on this point, but I need not decide the
issue, because it is ultimately irrelevant. As noted earlier,
Skinner squarely held that state action was present there even
in the absence of state compulsion. 

. And
for the reasons that I have explained, the same is true here. 13

13
   Although I thus do not reach the question of whether compulsion has
been shown here, I note parenthetically that I am also not sure that the
majority is correct in suggesting that, if compulsion had been
established, Meta would not be a proper defendant for such a claim. See
Opin. at 23. It may perhaps be true that the government-compelled
private party is not the proper defendant in a suit for damages or in a suit
challenging “governmental compulsion in the form of a generally
applicable law.” Sutton v. Providence St. Joseph Med. Ctr., 

 (9th Cir. 1999). And I recognize that the distributor was not
named as a defendant in the suit for injunctive relief in Bantam Books.
See Bantam Books, 372 U.S. at 60–61 (describing procedural history);
see also Bantam Books, Inc. v. Sullivan, 

 (R.I. 1961)
(noting that “[t]he distributor did not object to the commission’s action
and is not a party to the instant proceedings”). But I am not sure that the
         CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.                  81


    Likewise, it does not matter whether the majority is
correct in contending that the allegations here would not
suffice to establish state action under the traditional “joint
action test,” Lugar, 

. That test, according to
the majority, requires a showing that Meta agreed to take a
“specific action on the government’s say-so.” See Opin. at
18 (emphasis added). Once again, the point is ultimately
irrelevant. In Skinner, there was no such alleged agreement
to violate any specific person’s “rights in particular,” see
Opin. at 16 (citation omitted), and yet the Court found that
state action was present. At most, the majority has
established that particular alternative formulations of the
state-action test, which were developed with different
contexts in mind, are ill-suited to the unique circumstances
presented here. That calls, as in Skinner, for a more tailored
inquiry into whether, in light of those unique circumstances,
state action is nonetheless present. As in Skinner, it is
present here.
    The majority also raises a broader concern that a finding
of state action here would interfere with Meta’s exercise of
its own independent judgment over its platforms. Opin. at
18–19. Given that Meta may happen to share the
Government’s views that anti-vaccine speech and speakers
should be limited or blocked on its platforms, the majority


same conclusion follows if the compulsion test is applied in the unique
context presented here, i.e., a suit for injunctive relief against a private
party who, while exercising a government-granted ability to engage in
mass censorship, is allegedly the subject of particularized coercive
tactics from the Government. An injunction aimed at keeping the
Government’s coercive efforts away from such dangerous levers might
conceivably be addressed either to the target of those efforts (thus
counteracting them) or to the Government (or both). But, like the
majority, I need not ultimately decide this issue.
82     CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


argues that Meta should not be disabled from implementing
“those views simply because they happen to be shared by the
government.” See Opin. at 22. According to the majority,
“Meta has a First Amendment right” to censor any speech
on its platform with which it disagrees, see Opin. at 22
(emphasis added), and that it is solely up to Meta “to decide
what, if any, limits should apply to speech on those
platforms,” see Opin. at 31. Indeed, Meta contends—and the
majority appears to agree—that, under the Supreme Court’s
recent decision in Moody v. NetChoice, LLC, 

(2024), “all of the actions challenged here are protected
under the First Amendment” (emphasis added). In light of
these considerations, the majority suggests that the various
state-action tests should be narrowly construed so as to
preserve Meta’s asserted First-Amendment-based right to
freely censor speech on its platform. These arguments rest,
in my view, on an overstated view of Meta’s relevant First
Amendment rights, which do not give Meta an unbounded
freedom to work with the Government in suppressing speech
on its platforms.
    It may well be true that an ordinary publisher or
distributor would have a First Amendment right to
acquiesce, if persuaded, in governmental requests not to
publish or distribute particular works or speakers. The Court
in Bantam Books put loadbearing weight on the fact that the
Rhode Island courts had specifically found that the
distributor’s acquiescence in that case “was not voluntary.”

. It is therefore plausible to suppose that
Bantam Books might have come out differently if the
distributor had instead stated that it was persuaded by the
state commission’s views concerning what materials were
worth distributing and that, agreeing with those views, the
distributor affirmatively did not wish to promote the
        CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.       83


particular works at issue. Likewise, if a newspaper
affirmatively chooses to be, in effect, a mouthpiece for a
particular government or a particular official it supports, it
may well have an absolute First Amendment right to do so.
And to that extent, the majority would perhaps be correct in
suggesting that a newspaper’s constitutional right to opt to
kill whatever article it wants cannot be overcome by
relabeling, as “joint action,” the newspaper’s discussions
with government officials over whether to bury a story.
    But it does not follow from any of this that Meta has the
exact same scope of constitutional freedom with respect to
the speech of others on its mega-platforms. As I have
repeatedly explained, when it comes to the operation of the
sort of platforms at issue here, Meta simply does not occupy
the same position as a traditional newspaper publisher or a
book distributor. Rather, because its ability to operate its
massive platform rests dispositively on the immunity
granted as a matter of legislative grace in § 230, Meta is a bit
of a novel legal chimera: it has the immunity of a conduit
with respect to third-party speech, based precisely on the
overriding legal premise that it is not a publisher; its
platforms’ massive scale and general availability to the
public further make Meta resemble a conduit more than any
sort of publisher; but Meta has, as a practical matter, a
statutory freedom to suppress or delete any third-party
speech while remaining liable only for its own affirmative
speech. And, as the Supreme Court recently recognized,
Meta is engaged in expressive activity protected by the First
Amendment when it “curat[es]” Facebook’s “NewsFeed” in
a way that “create[s] a distinctive expressive offering.”
Moody, 144 S. Ct. at 2405. But I am aware of no historical
precedent that meaningfully corresponds to such a hybrid
entity, and I do not think we should simply assume that it has
84      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


exactly the same constitutional rights with respect to third-
party speech on its platforms as a newspaper publisher, a
book distributor, or a parade organizer. Moody did not
address the precise scope of Meta’s First Amendment rights
over its platform, see id. at 2407 (finding it unnecessary to
resolve what level of scrutiny applied to the restrictions at
issue there), and Moody did not confront or decide any
question as to whether Meta has an absolute constitutional
right to coordinate with the Government to suppress third-
party speech on its platforms.
    We likewise need not, and should not, decide in this case
exactly what degree of First Amendment protection, if any,
Meta has with respect to working with the Government to
censor particular viewpoints or speakers on its platforms. It
suffices for purposes of this case to note that the mega-
platforms at issue here differ from traditional publishers or
distributors in a critical respect that is directly relevant to the
state-action question and that, in my view, warrants a
different result—and that does so regardless of Meta’s own
invocation of the First Amendment. As I have explained,
Meta would be better positioned to argue for the full
constitutional freedom of a traditional publisher—including
the freedom to agree to, and implement, the Government’s
censorship preferences—if it operated its website in all
respects like a traditional publisher by individually
reviewing, selecting, and limiting exactly what third-party
speech it will publish. In such a circumstance, it would
happen to have § 230 immunity, but (as with a newspaper)
that immunity would not be essential to its very existence or
ability to operate its platforms.
    But in critical reliance on the Government’s creation of
an immunized censorship power, Meta instead chose to scale
up its operations in a way that has produced gigantic
       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.      85


platforms that comprise a unique assemblage of features that
make it part conduit, part distributor, and part publisher.
This central fact makes a difference. That is, I do not think
that Meta’s critical reliance on the government-created
ability to engage in mass censorship is a factor that can
properly be ignored either in the state-action inquiry or in
assessing whether, like the above-described acquiescing
newspaper publisher, Meta has, so to speak, a constitutional
right to suppress third-party speech that the Government
“persuades” it to censor. Although Meta’s operational
reliance on § 230’s immunity is not alone enough to render
Meta a state actor, that factor contributes positively towards
a finding of state action when combined with other
considerations. In particular, with this critical factor in
place, if Meta then affirmatively engages with the
Government as to how to exercise its government-granted
authority in order to widely suppress particular subjects or
speakers on its mega-platforms, that additional element
suffices to cross over the state-action line and to implicate
the First Amendment’s protections with respect to the
targeted speakers. And for that reason, I perceive no basis
for concluding that Meta, in operating such unprecedented
legally-hybrid platforms, has any sort of supervening
constitutional right to team up with the Government to
suppress the speech of particular speakers, or on particular
topics, on such immunized mega-platforms.
    The majority worries that treating Meta as a state actor
here would contravene the underlying purpose of the state-
action doctrine, which is to “protect[] a robust sphere of
individual liberty” within which private actors may operate.
See Opin. at 12 (quoting Halleck, 587 U.S. at 808). But in
this distinctive scenario, applying the state-action doctrine
promotes individual liberty by keeping the Government’s
86      CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


hands away from the tempting levers of censorship on these
vast platforms. To be sure, it means that Meta does not have
the “liberty” to work together with the Government in
deciding how to suppress the speech of millions of people,
but Meta otherwise retains its full authority to operate its
platform within the bounds of the law. A contrary rule
would mean that the Government can create a special
immunized power for private entities to suppress speech on
a mass scale and then request and receive, from those private
entities, an ability to influence the exercise of those levers of
censorship. That would thwart the First Amendment’s core
purpose to “prevent[] the government from tilting public
debate in a preferred direction.” Moody, 144 S. Ct. at 2407
(simplified).
    The majority suggests that finding state action here
would produce a parade of horribles, because it would
supposedly hamper the Government’s ability to work with
platform operators to restrict minors’ access to pornographic
speech or to address other types of speech as to which the
Government has legitimate concerns. See Opin. at 30–31.
But saying that the First Amendment is implicated is not the
same as saying that it is violated. Where the category of
speech at issue is either unprotected (e.g., child pornography,
fraudulent advertising) or is otherwise subject to legitimate
direct regulation by the Government, see Reno v. ACLU, 

 (1997) (reaffirming that the Government has
“‘a compelling interest in protecting the physical and
psychological well-being of minors’ which extend[s] to
shielding them from indecent messages that are not obscene
by adult standards” (citation omitted)), or where the
Government’s interest involves, for example, malign foreign
actors operating outside the United States, see Agency for
Int’l Dev. v. Alliance for Open Soc’y Int’l, Inc., 591 U.S.
       CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.    87


430, 439 (2020) (holding that “foreign organizations
operating abroad do not possess constitutional rights”), the
Government may also properly seek to achieve its legitimate
ends indirectly, through consultation with operators of
mega-platforms. What allegedly occurred here, however, is
quite different, because Meta and the Government worked
cooperatively together to suppress the concededly truthful
speech of Americans concerning vaccines, and the
Government sought to do so for the illegitimate purpose of
dampening opposition to the Government’s preferred
vaccine policies. See Rosenberger v. Rector & Visitors of
Univ. of Va., 

 (1995) (“The government
must abstain from regulating speech when the specific
motivating ideology or the opinion or perspective of the
speaker is the rationale for the restriction.”). Here, it is
alleged, the Government worked with “private persons to
accomplish what it is constitutionally forbidden to
accomplish.” Norwood v. Harrison, 

(1973) (citation omitted); see also National Rifle Ass’n v.
Vullo, 

 (2024) (stating that “a government
official cannot do indirectly what she is barred from doing
directly”).
                             V
    I concur in the majority’s opinion to the extent that it
upholds the district court’s dismissal of CHD’s Lanham Act
claim, its RICO claim, and its claims against additional
Defendant Science Feedback. For the reasons I have
explained, I would affirm the dismissal of the Takings
Clause claim; the dismissal of the First Amendment claims
against Zuckerberg and the Poynter Institute; and the Bivens
First Amendment claim for monetary damages against Meta.
But I would reverse as to CHD’s First Amendment claim for
88     CHILDREN’S HEALTH DEF. V. META PLATFORMS, INC.


injunctive and declaratory relief against Meta, and to that
extent, I respectfully dissent.