Opinion
No. 23-5110
07-30-2024
PEOPLE FOR THE ETHICAL TREATMENT OF ANIMALS, et al., Appellants v. Lawrence A. TABAK, in his official capacity as Acting Director of the National Institutes of Health and Xavier Becerra, in his official capacity as Secretary of the U.S. Department of Health and Human Services, Appellees
Stephanie Krent argued the cause for appellants. With her on the briefs were Ashley Ridgway, Katherine A. Fallow, Alexia Ramirez, and Jameel Jaffer. Sophia Cope and David Greene were on the brief for amici curiae Electronic Frontier Foundation and Foundation for Individual Rights and Expression in support of appellants. Jennifer L. Utrecht, Attorney, U.S. Department of Justice, argued the cause for appellees. With her on the brief were Brian M. Boynton, Principal Deputy Assistant Attorney General, and Daniel Tenny, Attorney.
Appeal from the United States District Court for the District of Columbia (No. 1:21-cv-02380) Stephanie Krent argued the cause for appellants. With her on the briefs were Ashley Ridgway, Katherine A. Fallow, Alexia Ramirez, and Jameel Jaffer. Sophia Cope and David Greene were on the brief for amici curiae Electronic Frontier Foundation and Foundation for Individual Rights and Expression in support of appellants. Jennifer L. Utrecht, Attorney, U.S. Department of Justice, argued the cause for appellees. With her on the brief were Brian M. Boynton, Principal Deputy Assistant Attorney General, and Daniel Tenny, Attorney. Before: Henderson, Millett, and Garcia, Circuit Judges. Garcia, Circuit Judge:
Appellants are the nonprofit People for the Ethical Treatment of Animals ("PETA") and two animal rights advocates, Madeline Krasno and Ryan Hartkopf, who use social media to advocate against animal testing. They frequently commented on the official Facebook and Instagram pages of appellee National Institutes of Health ("NIH"), criticizing NIH's funding of research conducted on animals. Those efforts ran headfirst into NIH's social media moderation policy, which prohibits, as relevant here, "off-topic posts." To enforce this policy, NIH deployed keyword filters—which automatically hide all comments with the chosen keywords—to filter out comments containing words that frequently appeared in posts that it considered "off-topic," such as the terms "animal," "testing," and "cruel." Appellants' and all other users' comments containing those words were thus filtered out and not viewable to the public.
Appellants argue that NIH's policy violates the First Amendment. We must decide what type of forum NIH's comment threads are and whether NIH's social media moderation policy, as implemented through its keyword filters, is constitutional. The district court held that the comment threads were limited public forums and upheld NIH's speech restrictions as reasonable.
We agree that NIH's comment threads are limited public forums because the government has signaled its intent to limit the discussion on those threads to specific subjects. But we hold that NIH's "off-topic" restriction, as implemented through its keyword filters, is not reasonable in light of the purpose of the forum and is therefore unconstitutional under the First Amendment.
I
A
The parties stipulated to the following facts. See Joint Stipulation of Facts ("Jt. Stip."), People for the Ethical Treatment of Animals v. Tabak ("PETA"), No. 21-cv-2380 (D.D.C. Feb. 11, 2022), ECF No. 28. NIH is the primary federal agency charged with performing and supporting biomedical and behavioral research. It maintains verified Facebook and Instagram pages to "communicate and interact with citizens" about agency-related work. Jt. Stip. ¶ 36. NIH often posts highlights of NIH-funded researchers and their research, interviews with experts, and news updates. Posts are viewable by the public, and Facebook and Instagram users can comment on them.
NIH's Comment Guidelines govern comments on NIH's social media. See NIH Comment Guidelines, (last updated March 13, 2019), https://perma.cc/YN6J-8GFH. Those publicly available guidelines "encourage" the public to "share [their] thoughts and ideas" and state that "NIH blogs are not intended to serve as public forums." Id. To "encourage respectful and constructive dialogue," the Guidelines prohibit "vulgar, obscene, profane, threatening, or abusive language," "discriminatory language," "endorsements of commercial products," "repetitive posts," "spam or undecipherable language," "links to external sites," and, most relevant here, "off-topic posts." Id. (capitalization modified). The Guidelines also ask that "comments be respectful and relevant to the specific topic." Id.
Facebook and Instagram accountholders—such as NIH here—have at their disposal several tools to moderate comments. Accountholders can block individual users and manually delete or hide comments. They can also turn on the platforms' default filters, which hide from public view comments that contain profanity or other offensive words. On Facebook, the hidden comment is then viewable only by the user and their friends. On Instagram, users may view hidden comments by scrolling to the bottom of a comment thread and clicking "view hidden comments." NIH enabled the platform's default filters here.
Accountholders may also use custom keyword filters—the primary focus of this case. An accountholder can specify a custom list of keywords to be filtered out; any past or future comment containing those keywords is then automatically hidden. On Facebook, a comment hidden by custom keyword filters is automatically hidden from public view but remains visible to the user who posted the comment and their friends. On Instagram, a comment hidden by custom keyword filters is visible only to the user and the account holder if they choose to "view hidden comments." Typically, the user who posted the filtered-out comment is not notified. NIH created custom keyword filters to implement its Commenting Guidelines.
As of the date of the complaint, September 9, 2021, NIH's Facebook keyword filters consisted of the following words: "PETA, PETALatino, animal(s), animales, animalitos, cats, gatos, chimpanzee(s), chimp(s), hamster(s), marmoset(s), monkey(s), monkies, mouse, mice, primate(s), sex experiments, cruel, cruelty, revolting, torment(ing), torture(s), torturing, #believemothers, marijuana, cannabis, Hitler, nazi," as well as the names of two researchers who have conducted experiments on monkeys ("Suomi," "Harlow"), various URL parts (e.g., www., gmail, .org, .com), and expletives. Jt. Stip. ¶ 58 (capitalization modified); see Compl., Attachment 1. On Instagram, the following words were filtered: "PETA, #stopanimaltesting, #stoptesting, #stoptestingonanimals, animal(s), chimpanzee(s), chimps, monkey(s), experiment, hurt(ing), kill, stop, test(ing), testing facility, tortur(ing), pedos, rapist," as well as four emojis and an expletive. Jt. Stip. ¶ 58 (capitalization modified); see Compl., Attachment 2. NIH said it added these terms "to target comments frequently made on NIH's social media pages that the NIH believes would violate its comment moderation guidelines." Jt. Stip. ¶ 58. While NIH maintains it has the "authority to manually review any and all comments before and after they are posted," "in practice" NIH implemented its guidelines only through use of keyword filters. Id. ¶ 61.
B
On September 9, 2021, PETA, Krasno, and Hartkopf sued NIH in district court, alleging that NIH's use of keyword filters violated their First Amendment rights. PETA is a nonprofit organization dedicated to advocating for animal rights, and the individual appellants are animal rights advocates. Appellants use social media to raise awareness for their causes, which include opposing animal testing and animal mistreatment in laboratories. In particular, appellants frequently comment on NIH's social media pages to protest NIH's funding of animal testing. Many of their comments have been automatically filtered out by NIH's keyword filters. For example, Krasno attempted to comment the following on an NIH Instagram post depicting a cell infected with COVID-19: "It's time we had an open conversation about all the animal testing you fund. What a waste of life and resources." Jt. Stip. ¶ 73. That comment was automatically hidden from public view because "animal" and "testing" were on NIH's keyword filters list. Id.
In December 2021, NIH removed "PETA" and "PETALatino" from its keyword filters on both platforms and removed "#stopanimaltesting" and "#stoptesting" from its Instagram filters. Jt. Stip. ¶ 60. The parties cross-moved for summary judgment. The district court granted summary judgment to the government, holding that NIH's keyword filters were viewpoint-neutral and reasonable restrictions in a limited public forum. See PETA v. Tabak, 2023 WL 2809867, at *14 (D.D.C. Mar. 31, 2023). This appeal followed.
Appellants' complaint also named the Department of Health and Human Services ("HHS") as a defendant, but the district court's order did not resolve the claims against HHS. This raised the question whether the district court's order is final and ready for our review under 28 U.S.C. § 1291. To resolve this issue, appellants, with the government's consent, filed a motion to dismiss their claims against HHS in district court, which the district court granted. Minute Order, PETA, No. 21-cv-2380 (D.D.C. Jan. 11, 2024).
II
We review the district court's grant of summary judgment in favor of the government de novo, Bauer v. Fed. Deposit Ins., 38 F.4th 1114, 1121 (D.C. Cir. 2022), and have an "obligation to 'make an independent examination of the whole record,' " Liberty Lobby, Inc. v. Rees, 852 F.2d 595, 598 (D.C. Cir. 1988) (quoting Bose Corp. v. Consumers Union of United States, Inc., 466 U.S. 485, 499, 104 S.Ct. 1949, 80 L.Ed.2d 502 (1984)).
A
The speech restrictions challenged here apply only to NIH's comment threads, which the parties agree are government-controlled property. See Appellants' Brief 33; Appellee's Brief 26. On government-controlled property, the standard by which speech restrictions are evaluated "depends on the nature of the relevant forum." Cornelius v. NAACP Legal Def. & Educ. Fund, Inc., 473 U.S. 788, 800, 105 S.Ct. 3439, 87 L.Ed.2d 567 (1985).
Traditional public forums—like parks and sidewalks—are "places which by long tradition . . . have been devoted to assembly and debate." Perry Educ. Ass'n v. Perry Loc. Educators' Ass'n, 460 U.S. 37, 45, 103 S.Ct. 948, 74 L.Ed.2d 794 (1983). In such forums, the government may "impose reasonable time, place, and manner restrictions on private speech, but restrictions based on content must satisfy strict scrutiny, and those based on viewpoint are prohibited." Minn. Voters All. v. Mansky, 585 U.S. 1, 11, 138 S.Ct. 1876, 201 L.Ed.2d 201 (2018). That same strict standard applies to a designated public forum, which is a space that has " 'not traditionally been regarded as a public forum' but which the government has 'intentionally opened up for that purpose.' " Id. (quoting Pleasant Grove City v. Summum, 555 U.S. 460, 469, 129 S.Ct. 1125, 172 L.Ed.2d 853 (2009)).
In contrast, a nonpublic forum, such as a museum or office, is a place "not by tradition or designation" a public forum and one in which the government may restrict speech so long as the "regulation on speech is reasonable and not an effort to suppress expression merely because public officials oppose the speaker's view." Perry, 460 U.S. at 46, 103 S.Ct. 948. Finally, what we have sometimes termed a "hybrid case" is a "limited public forum, in which the Government has 'create[d] a forum that is limited to use by certain groups or dedicated solely to the discussion of certain subjects.' " Price v. Garland, 45 F.4th 1059, 1068 (D.C. Cir. 2022) (quoting Pleasant Grove, 555 U.S. at 470, 129 S.Ct. 1125) (alteration in original). As in nonpublic forums, speech restrictions in limited forums need only be viewpoint neutral and "reasonable in light of the purpose served by the forum." Good News Club v. Milford Cent. Sch., 533 U.S. 98, 106-07, 121 S.Ct. 2093, 150 L.Ed.2d 151 (2001) (quoting Cornelius, 473 U.S. at 806, 105 S.Ct. 3439).
Appellants argue that NIH's comment threads are designated public forums; the government contends they are limited forums. The district court agreed with the government on this threshold question, and so do we.
"The touchstone for determining whether government property is a designated public forum [or instead a nonpublic or limited forum] is the government's intent in establishing and maintaining the property." Stewart v. D.C. Armory Bd., 863 F.2d 1013, 1016 (D.C. Cir. 1988). The government does not create a designated public forum "by inaction or by permitting limited discourse" but rather "only by intentionally opening a nontraditional forum for public discourse." Cornelius, 473 U.S. at 802, 105 S.Ct. 3439. Put otherwise, "the government must make an affirmative choice to open up its property for use as a public forum." United States v. Am. Libr. Ass'n, 539 U.S. 194, 206, 123 S.Ct. 2297, 156 L.Ed.2d 221 (2003) (plurality opinion) (emphasis added). To discern the government's intent, we look at the government's "stated purpose" and other " 'objective indicia of intent' such as 'the nature of the property, its compatibility with expressive activity, and the consistent policy and practice of the government.' " Bryant v. Gates, 532 F.3d 888, 896 (D.C. Cir. 2008) (quoting Stewart, 863 F.2d at 1016-17) (emphasis omitted).
Here, those factors demonstrate that NIH sought to limit the forum "solely to the discussion of certain subjects" and that it is therefore a limited public forum. Pleasant Grove, 555 U.S. at 470, 129 S.Ct. 1125.
To begin, the government stated in its Comment Guidelines that its "blogs are not intended to serve as public forums." NIH Comment Guidelines, supra. It further stated that the purpose of its social media pages is to "disseminate health information by communicating important public health information" and "engaging the public for educational purposes about public health," Jt. Stip. ¶ 87, which suggests NIH was limiting the pages "solely to the discussion of certain subjects," Pleasant Grove, 555 U.S. at 470, 129 S.Ct. 1125, namely those relevant to NIH's public health work.
NIH also "prospectively and categorically set subject matter regulations" to effectuate those stated aims. Archdiocese of Wash. v. Wash. Metro. Area Transit Auth., 897 F.3d 314, 324 (D.C. Cir. 2018). NIH published online its Comment Guidelines, which prohibit several categories of speech including off-topic posts; endorsements of commercial products; and profane, repetitive, or discriminatory comments. See NIH Comment Guidelines, supra; see also id. ("We ask that your comments be respectful and relevant to the specific topic."). Moreover, as detailed further below, NIH sought to enforce those guidelines by using a range of keyword filters. The guidelines and use of keyword filters indicate that the comment sections are limited to discussion of certain subjects rather than open "for use by the public as a place for expressive activity" in general. Perry, 460 U.S. at 45, 103 S.Ct. 948.
Several courts have held that when the government indicates its intent to limit discussion to certain subject matter by setting restrictions in analogous situations, the forum is a limited public forum. One common example is when a school board limits public comment during its meetings to only those topics on its agenda. See Barrett v. Walker Cnty. Sch. Dist., 872 F.3d 1209, 1225 (11th Cir. 2017) (school board meeting was limited forum in part because the board "limits discussion to certain topics"); see also Steinburg v. Chesterfield Cnty. Plan. Comm'n, 527 F.3d 377, 385 (4th Cir. 2008) (similar); Galena v. Leone, 638 F.3d 186, 199 (3d Cir. 2011) (similar).
In the social media context, courts have also found that the comment threads of government social media pages are designated public forums when the pages are open for comment without restrictions and limited public forums when the government prospectively sets restrictions. See, e.g., Garnier v. O'Connor-Ratcliff, 41 F.4th 1158, 1178-79 (9th Cir. 2022) (pages were designated public forums when "available to the public without any restriction on the form or content of comments" but became limited forums when the government used word filters and other speech restrictions), vacated and remanded on other grounds, 601 U.S. 205, 144 S.Ct. 717, 218 L.Ed.2d 138 (2024); Davison v. Randall, 912 F.3d 666, 682 (4th Cir. 2019) (comment section held to be designated public forum when the government "placed no restrictions" on use and access to the page), as amended (Jan. 9, 2019); see also Knight First Amend. Inst. at Columbia Univ. v. Trump, 928 F.3d 226, 237 (2d Cir. 2019) (similar), judgment vacated on other grounds, — U.S. —, 141 S. Ct. 1220, 209 L.Ed.2d 519 (2021).
Despite those stated policies, and to distinguish the cases just discussed, appellants argue that NIH's enforcement of those restrictions was inconsistent to an extent that shows NIH's true intent was to make the comment threads designated public forums and block only animal advocacy speech. Appellants' Brief 35-39. As support, appellants point to cases holding that the government created designated public forums when it announced speech restrictions but then ignored those restrictions in practice. See, e.g., Grace Bible Fellowship, Inc. v. Me. Sch. Admin. Dist. No. 5, 941 F.2d 45, 47 (1st Cir. 1991); Hopper v. City of Pasco, 241 F.3d 1067, 1081 (9th Cir. 2001); see also Lebron v. Wash. Metro. Area Transit Auth., 749 F.2d 893, 896 (D.C. Cir. 1984). And as proof of inconsistent enforcement, appellants emphasize that a substantial number of comments that seemingly violate NIH's guidelines remained on NIH's posts. See, e.g., Jt. Stip. Ex. 56 at 4.
We disagree. This is not a case in which the government made either no effort, or a minimal one, to enforce its stated policy. Rather, NIH attempted to implement its stated policies through keyword filters, which apply across the whole page and filter all comments containing those words. See Jt. Stip. ¶ 30. Those filters captured several content areas other than the challenged animal-testing-related filters, such as profanity, drug-related words, #believemothers, several emojis, offensive words, and terms related to external links. See Compl., Attachment 1 (Facebook filters); Compl., Attachment 2 (Instagram filters). NIH also enforced its rules by enabling the default Facebook and Instagram filters that restrict offensive language. Jt. Stip. ¶¶ 56-57.
It is true that a number of comments that seemingly violate NIH's stated policy remained on NIH posts, but that fact has less salience here than did similar facts in the cases on which appellants rely. For example, in Hopper, the Ninth Circuit found that a city-owned art exhibit, despite having a stated policy against controversial works, was a designated public forum because the city had never enforced that policy until it excluded the plaintiffs' works. 241 F.3d at 1071, 1078. Importantly, the city signed a contract with each individual artist to display their work but had never before applied its stated policy. Id. at 1072. Those facts supported the inference that the government did not truly intend to enforce that policy and instead created a public forum. Id. at 1078, 1080; see also Grace Bible Fellowship, 941 F.2d at 46 (plaintiffs sought and were denied a lease to use school facilities for religious purposes when other speakers using facilities for expressive purposes were approved); Lebron, 749 F.2d at 895 & n.6 (plaintiff sought and was denied approval for a political bus advertisement when prior political advertisements were approved).
Here, by contrast, NIH did not affirmatively decide not to enforce its guidelines, nor did it approve the specific comments that violated its guidelines. Rather, the record shows that because NIH lacked "the resources to manually" review each of the innumerable posted comments, it attempted to prospectively block—through keyword filters—a variety of comments that it believed would violate its guidelines. See Compl., Attachment 1 (Facebook filters); Compl., Attachment 2 (Instagram filters). That enforcement was certainly imperfect, and some speech that contravenes NIH's policies was not blocked or removed. But that fact is far less relevant here because, given the context, it does not reflect an "affirmative choice" by NIH to allow the violative comments. Am. Libr. Ass'n, 539 U.S. at 206, 123 S.Ct. 2297. As the district court aptly put it, a court's "examination of the government's actual practice in enforcing its policy is not a gotcha game; rather it is a tool to smoke out the government's 'intent' with regard to the forum's purpose—the 'touchstone' of the forum inquiry." PETA, 2023 WL 2809867, at *9 (quoting Bryant, 532 F.3d at 895-96).
In short, because NIH attempted to remove a range of speech violating its policies, and because the existence of violative speech in the forum bears less on the government's intent here than it did in the appellants' cited cases, we find sufficient evidence that the government intended to limit the forum to only speech that meets its public guidelines.
This analysis echoes the Supreme Court's plurality opinion in American Library Association. See 539 U.S. at 208, 123 S.Ct. 2297. There, petitioners argued that public library computers were designated public forums because after the libraries made internet access available to patrons, they had restricted access to only one "inappropriate" subject (pornography) but had not made a judgment about whether the vast amount of remaining content available on the internet was appropriate. Id. at 206-07, 123 S.Ct. 2297. The Supreme Court held that this underinclusivity did not make the computers designated public forums. Id. at 208, 123 S.Ct. 2297; see id. at 206, 123 S.Ct. 2297. So too here. While NIH's filters certainly do not capture all types of off-topic or other comments that might violate NIH's Guidelines, they nonetheless still support finding that the government intended to keep its comment sections a limited public forum.
Finally, whether the forum is "compatible with expressive activity" factors into the forum analysis. Stewart, 863 F.2d at 1019. We recognize social media is inherently compatible with expressive activity given it is one of the "most important places . . . for the exchange of views." Packingham v. North Carolina, 582 U.S. 98, 104, 137 S.Ct. 1730, 198 L.Ed.2d 273 (2017). This factor, however, is not dispositive on its own. See Stewart, 863 F.2d at 1019 (noting that "compatibility provides some evidence of the government's intent"). After considering NIH's stated purpose, policy, and enforcement efforts, we conclude that NIH intended to create a forum limited "solely to the discussion of certain subjects." Pleasant Grove, 555 U.S. at 470, 129 S.Ct. 1125.
B
In a limited public forum, speech restrictions may be "based on subject matter" if the restrictions are "reasonable in light of the purpose served by the forum" and "viewpoint neutral." Cornelius, 473 U.S. at 806, 105 S.Ct. 3439. Appellants do not challenge the entirety of NIH's Comment Guidelines; they focused solely on the constitutionality of the off-topic restriction as implemented by the keyword filters. We therefore examine only that restriction.
Reasonableness is to be assessed in light of the purpose of the forum, which here is to "communicate and interact with citizens," Jt. Stip. ¶ 36, and to "encourage respectful and constructive dialogue" through the public's comments, NIH Comment Guidelines, supra. Reasonableness in this context is thus necessarily a more demanding test than in forums that have a primary purpose that is less compatible with expressive activity, like the football stadium "forum" in Stewart. See 863 F.2d at 1019-20 (noting that a forum's compatibility with expressive activity factors into both forum and reasonableness analysis). In service of those purposes, NIH's off-topic restriction furthers the "permissible objective[s]," Mansky, 585 U.S. at 13, 138 S.Ct. 1876, of creating comment threads dedicated to each post's topic and allowing the public to engage on that topic, instead of being distracted or overwhelmed by off-topic comments.
But NIH must "draw a reasonable line," id. at 16, 138 S.Ct. 1876, informed by "objective, workable standards," id. at 21, 138 S.Ct. 1876, between what is considered on-topic and what is considered off-topic. "Although there is no requirement of narrow tailoring," the government "must be able to articulate some sensible basis for distinguishing what may come in from what must stay out." Id. at 16, 138 S.Ct. 1876. This NIH has not done.
In the context of NIH's posts— which often feature research conducted using animal experiments or researchers who have conducted such experiments—to consider words related to animal testing categorically "off-topic" does not "ring[ ] of common-sense." United States v. Kokinda, 497 U.S. 720, 734, 110 S.Ct. 3115, 111 L.Ed.2d 571 (1990) (plurality opinion) (internal quotation marks omitted). For example, consider NIH's July 20, 2021 Instagram post, which featured a photo of the eye of a zebrafish. Jt. Stip., Ex. 9. The caption read, in part: "This picture of an anesthetized adult zebrafish was taken with a powerful microscope that uses lasers to illuminate the fish." Id. It is unreasonable to think that comments related to animal testing are off-topic for such a post. Yet a comment like "animal testing on zebrafish is cruel" would have been filtered out because "animal," "testing," and "cruel" are all blocked by NIH's keyword filters.
The government admits that animal testing comments would be on-topic for that post and instead argues that the off-topic rule is still reasonable because a reasonable policy may be both over- and underinclusive. See Oral Argument Tr. 31:1-12. That argument assumes the zebrafish post is an outlier. But the record indicates otherwise. A substantial portion of the NIH posts included in the stipulated record either directly depict animals or discuss research conducted on animals. See Jt. Stip. Exs. 8-9 (posts depicting animals); id., Exs. 6, 11, 32, 41, 52, 56-57 (posts discussing research conducted on animals); see also Appellants' Brief 26-27 n.10. To say that comments related to animal testing are categorically off-topic when a significant portion of NIH's posts are about research conducted on animals defies common sense.
Worse, the government fails to provide any definition of "off-topic" in its Comment Guidelines, to its social media moderators, or even in this litigation. See Oral Arg. Tr. 29:4-7 (NIH arguing that "off topic" is a "commonly understood" term but providing no explicit definition); id. at 29:20-21 (NIH stating that "[t]here's nothing in the comment guidelines that define[s] what off topic means"); id. at 54:22-55:25 (NIH stating its moderators use their "experience"). And without such guidance, in this context at least, it is far from clear where the line between off-topic and on-topic lies.
Take another recurring example from the record: An NIH post highlighting a study by a researcher who regularly conducts experiments on animals but did not conduct any such experiments in the particular study highlighted. See Jt. Stip., Exs. 13, 17, 25, 29; Appellants' Brief 27 n.11. One could argue that a comment criticizing that researcher's general use of animal testing is on-topic, because the post introduced the researcher as a "topic" of the post. But one could also reasonably think that such a comment is off-topic because the specific study highlighted is the relevant "topic," and the study itself did not involve animal testing. Simply announcing a rule against "off-topic" comments does not provide "objective, workable standards," Mansky, 585 U.S. at 16, 138 S.Ct. 1876, to guide either NIH's social media moderators or the public as to how to divine "what may come in from what must stay out," id. at 21, 138 S.Ct. 1876. Though we have never required a speech restriction to demonstrate "perfect clarity," the problem with NIH's off-topic rule goes "beyond close calls on borderline or fanciful cases." Mansky, 585 U.S. at 21, 138 S.Ct. 1876 (internal quotation marks omitted). Moreover, while NIH claimed in this litigation that there was an "alarming number of repetitive, off-topic" comments about animal testing, Appellee's Brief 25, NIH provided no line (either to us or to its own social media moderators) demarcating what is an acceptable number of off-topic posts and what is too much.
"It is 'self-evident' that an indeterminate prohibition carries with it '[t]he opportunity for abuse, especially where [it] has received a virtually open-ended interpretation.' " Mansky, 585 U.S. at 21, 138 S.Ct. 1876 (quoting Bd. of Airport Comm'rs of L.A. v. Jews for Jesus, 482 U.S. 569, 576, 107 S.Ct. 2568, 96 L.Ed.2d 500 (1987)). It is perhaps no surprise then that NIH's moderators originally added terms like "PETA" and "#stopanimaltesting" to the keyword filters which were then, during this litigation, removed once NIH realized those terms "may have signaled a certain viewpoint." Defendants' Cross-Motion for Summary Judgment 26, PETA, No. 21-cv-2380, (D.D.C. May 6, 2022), ECF No. 31. The district court forgave these keyword choices as "an overzealous attempt by a NIH social media manager to tamp down irrelevant posts." PETA, 2023 WL 2809867, at *13 n.13. To us, however, these missteps are confirmation that NIH's policy does not "guide[ ]" its social media managers with any "objective, workable standards." Mansky, 585 U.S. at 21, 138 S.Ct. 1876. That undermines the reasonableness of the NIH policy.
NIH's off-topic policy, as implemented by the keywords, is further unreasonable because it is inflexible and unresponsive to context. In American Library Association, for example, even though the pornography filters erroneously blocked some websites that did not show pornographic content, the Supreme Court held that the policy was reasonable in part because library patrons could easily disable the filtering software by asking a librarian to unblock the site either temporarily for their own use or permanently for use by others. 539 U.S. at 209, 123 S.Ct. 2297.
By contrast, NIH's moderation policy lacks comparable features. The keyword filters apply automatically to comments on all NIH posts. They do not account for the topic of any given post or the context in which a comment is made—for example, a long comment that is generally responsive to the post would be filtered out if it uses any one of the keywords. Further, NIH does not employ any manual review of comments to restore otherwise on-topic comments that have been removed, turn off its filters when it posts content that is likely to make certain keywords relevant, or even routinely review its keyword list to consider whether its keywords should be removed (at least absent a lawsuit). See Jt. Stip. ¶ 61; Oral Arg. Tr. 54:1-4. Users seemingly have little, if any, ability to ask NIH to restore their comments; indeed, they typically are not notified when their comments are filtered out. See Jt. Stip. ¶¶ 18, 34. The permanent and context-insensitive nature of NIH's speech restriction reinforces its unreasonableness, especially absent record evidence that comments about animal testing materially disrupt NIH's ability to meet its objective of communicating with citizens about NIH's work.
Finally, NIH's off-topic restriction is further compromised by the fact that NIH chose to moderate its comment threads in a way that skews sharply against the appellants' viewpoint that the agency should stop funding animal testing by filtering terms such as "torture" and "cruel," not to mention terms previously included such as "PETA" and "#stopanimaltesting." The right to "praise or criticize governmental agents" lies at the heart of the First Amendment's protections, Mills v. Alabama, 384 U.S. 214, 219, 86 S.Ct. 1434, 16 L.Ed.2d 484 (1966), and censoring speech that contains words more likely to be used by animal rights advocates has the potential to distort public discourse over NIH's work. The government should tread carefully when enforcing any speech restriction to ensure it is not viewpoint discriminatory and does not inappropriately censor criticism or exposure of governmental actions. See Ridley v. Mass. Bay Transp. Auth., 390 F.3d 65, 86 (1st Cir. 2004).
For all of these reasons, we hold that NIH's off-topic restriction, as currently presented, is unreasonable under the First Amendment. We therefore do not separately address whether the specific keywords used to implement the off-topic rule are, by themselves, viewpoint discriminatory.
III
The judgment of the district court is reversed, and we direct entry of summary judgment in favor of the appellants.
So ordered.