Court Makes It Clear: Government Submissions To Twitter Flagging Program Do Not Violate The 1st Amendment

There’s been a lot of discussion of late, especially because of the various Twitter Files, regarding where the line is between governments simply flagging content for social media websites to vet against their own policies as compared to unconstitutional and impermissible suppression of speech in violation of the 1st Amendment.

As we’ve highlighted over and over again, the courts of these so-called “jawboning” cases have been pretty clear that there needs to be a coercive element to make it a 1st Amendment violation. Judge Posner’s ruling in the Backpage v. Dart case in the 7th Circuit lays it out pretty clearly back in 2015, citing back to the 2nd Circuit’s Okwedy v. Molinari case:

The difference between government expression and intimidation—the first permitted by the First Amendment, the latter forbidden by it—is well explained in Okwedy v. Molinari, 333 F.3d 339, 344 (2d Cir. 2003) (per curiam): “the fact that a public-official defendant lacks direct regulatory or decisionmaking authority over a plaintiff, or a third party that is publishing or otherwise disseminating the plaintiff’s message, is not necessarily dispositive … . What matters is the distinction between attempts to convince and attempts to coerce. A public-official defendant who threatens to employ coercive state power to stifle protected speech violates a plaintiff’s First Amendment rights, regardless of whether the threatened punishment comes in the form of the use (or, misuse) of the defendant’s direct regulatory or decision-making authority over the plaintiff, or in some less-direct form.”

A few years back, we highlighted what we thought was an interesting case regarding social media and jawboning by state officials brought by Shiva Ayyadurai. Despite Ayyadurai’s history of trying to destroy our own site, as well as some of the dubious claims that resulted in his own case, we thought he raised a potentially worthwhile 1st Amendment question regarding where the line was between “convince” and “coerce” when it came to state officials complaining to Twitter about taking down content. It wasn’t clear where things fell in that case, especially as the government official there admitted they were trying to get Shiva’s tweets taken down. For whatever reason (and there were a variety of procedural oddities in the case), Shiva dropped his case so we never got a direct ruling in that one.

However, a similar case was more recently filed in California (and we mentioned briefly), in which lawyer Rogan O’Handley, a 2020 election truther, lost his Twitter account for violating Twitter’s policies. It came out that his account was one that was flagged by the California Secretary of State’s Office as a “trusted” flagger. So O’Handley sued California’s Secretary of State, Shirley Weber, along with Twitter, and the National Association of Secretaries of State.

The lower court had dismissed the case, and now the 9th Circuit took up the appeal, which… upheld the lower court ruling and tosses out the case. In other words: state officials merely flagging content by itself is not a violation of the 1st Amendment.

As always in these kinds of disputes, the specifics matter. O’Handley made a tweet alleging election fraud in California:

Audit every California ballot

Election fraud is rampant nationwide and we all know California is one of the culprits

Do it to protect the integrity of that state’s elections

The Secretary of State’s office flagged that tweet to Twitter via its Partner Support Portal saying the following:

Hi, We wanted to flag this Twitter post: https://twitter.com/DC_Draino/status/12370 73866578096129 From user @DC_Draino. In this post user claims California of being a culprit of voter fraud, and ignores the fact that we do audit votes. This is a blatant disregard to how our voting process works and creates disinformation and distrust among the general public.

The lower court dismissed on a variety of grounds, noting that Twitter wasn’t a state actor, and that his being banned from Twitter was “not fairly traceable to the Secretary’s actions” among other things. Both were appealed.

The appeals court easily tosses the claims against Twitter noting (of course) that Twitter is not the government:

O’Handley’s claims falter at the first step. Twitter did not exercise a state-created right when it limited access to O’Handley’s posts or suspended his account. Twitter’s right to take those actions when enforcing its content-moderation policy was derived from its user agreement with O’Handley, not from any right conferred by the State. For that reason, O’Handley’s attempt to analogize the authority conferred by California Elections Code § 10.5 to the “procedural scheme” in Lugar is wholly unpersuasive. Id. at 941. Lugar involved a prejudgment attachment system, created by state law, that authorized private parties to sequester disputed property. Id. Section 10.5, by contrast, does not vest Twitter with any power and, under the terms of the user agreement to which O’Handley assented, no conferral of power by the State was necessary for Twitter to take the actions challenged here.

Nor did Twitter enforce a state-imposed rule when it limited access to O’Handley’s posts and suspended his account for “violating the Twitter Rules . . . about election integrity.” As the quoted message that Twitter sent to O’Handley makes clear, the company acted under the terms of its own rules, not under any provision of California law.

That’s pretty straightforward. Also, the 9th Circuit notes that it really doesn’t matter that most of the accounts flagged by the Secretary of State’s office were later pulled down:

That Twitter and Facebook allegedly removed 98 percent of the posts flagged by the OEC does not suggest that the companies ceded control over their content-moderation decisions to the State and thereby became the government’s private enforcers. It merely shows that these private and state actors were generally aligned in their missions to limit the spread of misleading election information. Such alignment does not transform private conduct into state action.

Correlation is not causation in legal form.

And then we start to get into the more meatier question of whether or not there was any coercion which (again) is the key to all of this (this is still part of the analysis regarding whether or not Twitter has been turned into a state actor). The Court recognizes that there is none here.

In this case, O’Handley has not satisfied the nexus test because he has not alleged facts plausibly suggesting that the OEC pressured Twitter into taking any action against him. Even if we accept O’Handley’s allegation that the OEC’s message was a specific request that Twitter remove his November 12th post, Twitter’s compliance with that request was purely optional. With no intimation that Twitter would suffer adverse consequences if it refused the request (or receive benefits if it complied), any decision that Twitter took in response was the result of its own independent judgment in enforcing its Civic Integrity Policy. As was true under the first step of the Lugar framework, the fact that Twitter complied with the vast majority of the OEC’s removal requests is immaterial. Twitter was free to agree with the OEC’s suggestions—or not. And just as Twitter could pay greater attention to what a trusted civil society group had to say, it was equally free to prioritize communications from state officials in its review process without being transformed into a state actor.

The court notes that basic information sharing between governments and private actors does not make the private actors into state actors.

The relationship between Twitter and the OEC more closely resembles the “consultation and information sharing” that we held did not rise to the level of joint action in Mathis, 75 F.3d at 504. In that case, PG&E decided to exclude one of its employees from its plant after conducting an undercover investigation in collaboration with a government narcotics task force. Id. at 501. The suspended employee then sued PG&E for violating his constitutional rights under a joint action theory. Id. We rejected his claim because, even though the task force engaged in consultation and information sharing during the investigation, the task force “wasn’t involved in the decision to exclude Mathis from the plant,” and the plaintiff “brought no evidence PG&E relied on direct or indirect support of state officials in making and carrying out its decision to exclude him.” Id. at 504.

The same is true here. The OEC reported to Twitter that it believed certain posts spread election misinformation, and Twitter then decided whether to take disciplinary action under the terms of its Civic Integrity Policy. O’Handley alleges no facts plausibly suggesting either that the OEC interjected itself into the company’s internal decisions to limit access to his tweets and suspend his account or that the State played any role in drafting Twitter’s Civic Integrity Policy. As in Mathis, this was an arm’s-length relationship, and Twitter never took its hands off the wheel.

As for the claims directly against the Secretary of State, the 9th Circuit does find that O’Handley has standing, but still rejects his claims. The key part, again, is that Twitter gets to make its own decisions and the fact that the Secretary of State’s office flagged the tweet in no way changes that:

Here, as discussed above, the complaint’s allegations do not plausibly support an inference that the OEC coerced Twitter into taking action against O’Handley. The OEC communicated with Twitter through the Partner Support Portal, which Twitter voluntarily created because it valued outside actors’ input. Twitter then decided how to respond to those actors’ recommendations independently, in conformity with the terms of its own content-moderation policy

O’Handley tried to argue (as I’ve seen others as well) that the mere fact that the information sharing was coming from the government creates implicit intimidation factors, but the court, correctly, notes that this is not how any of this works:

O’Handley argues that intimidation is implicit when an agency with regulatory authority requests that a private party take a particular action. This argument is flawed because the OEC’s mandate gives it no enforcement power over Twitter. See Cal. Elec. Code § 10.5. Regardless, the existence or absence of direct regulatory authority is “not necessarily dispositive.” Okwedy, 333 F.3d at 344. Agencies are permitted to communicate in a non-threatening manner with the entities they oversee without creating a constitutional violation. See, e.g., National Rifle Association of America v. Vullo, 49 F.4th 700, 714–19 (2d Cir. 2022).

The court also rejects the idea that this was “retaliation” for O’Handley’s speech, noting that it doesn’t match up with the standards there either:

The retaliation-based theory of liability fails as well. To state a retaliation claim, a plaintiff must show that: “(1) he engaged in constitutionally protected activity; (2) as a result, he was subjected to adverse action by the defendant that would chill a person of ordinary firmness from continuing to engage in the protected activity; and (3) there was a substantial causal relationship between the constitutionally protected activity and the adverse action.” Blair v. Bethel School District, 608 F.3d 540, 543 (9th Cir. 2010) (footnote omitted).

O’Handley’s claim falters on the second prong because he has not alleged that the OEC took any adverse action against him. “The most familiar adverse actions are exercise[s] of governmental power that are regulatory, proscriptive, or compulsory in nature and have the effect of punishing someone for his or her speech.” Id. at 544 (citation and internal quotation marks omitted). Flagging a post that potentially violates a private company’s contentmoderation policy does not fit this mold.Rather, it is a form of government speech that we have refused to construe as “adverse action” because doing so would prevent government officials from exercising their own First Amendment rights. See Mulligan v. Nichols, 835 F.3d 983, 988–89 (9th Cir. 2016). California has a strong interest in expressing its views on the integrity of its electoral process. The fact that the State chose to counteract what it saw as misinformation about the 2020 election by sharing its views directly with Twitter rather than by speaking out in public does not dilute its speech rights or transform permissible government speech into problematic adverse action. See Hammerhead Enterprises, Inc. v. Brezenoff, 707 F.2d 33, 39 (2d Cir. 1983).

There is nothing surprising or out of the ordinary in the result of this case. It matches just fine with a large number of earlier “jawboning” style cases, including ones cited above like Okwedy, Bantam Books, and Backpage. However, since many seem eager to ignore all of this precedent, and because the facts are slightly different regarding social media and trusted flagging programs, it’s nice to see a clean ruling on these points.

Once again, the thing that matters is whether or not there is coercion. There may be cases where these programs or efforts tip over into coercion: and we should be vigilant in watching out for those scenarios. But mere information sharing, absent any form of coercion, cannot be a 1st Amendment violation.