Court Quickly Rejects California’s Deepfake Law As Blatantly Unconstitutional

Free Speech

from the well-that-was-fast dept

Well, things sure move fast in this world of AI regulations. Just last week, we noted that California Governor Gavin Newsom signed a pretty obviously unconstitutional set of laws regarding the use of deepfake imagery around “election communications.” Then, just hours later, he was sued for it.

This week, federal judge John Mendez has already put one law on hold, AB 2839*, noting that it seems quite clearly to be unconstitutional. Oftentimes, these kinds of rulings on preliminary injunctions will talk about “likelihood of success” and highlight how it will probably be found unconstitutional if it goes through the full legal process.

However, here, Judge Mendez just said point blank that the law is unconstitutional.

AB 2839 does not pass constitutional scrutiny because the law does not use the least restrictive means available for advancing the State’s interest here. As Plaintiffs persuasively argue, counter speech is a less restrictive alternative to prohibiting videos such as those posted by Plaintiff, no matter how offensive or inappropriate someone may find them. ‘“Especially as to political speech, counter speech is the tried and true buffer and elixir,” not speech restriction.’….

While California has a valid interest in protecting the integrity and reliability of the electoral process, AB 2839 is unconstitutional because it lacks the narrow tailoring and least restrictive alternative that a content based law requires under strict scrutiny.

As you’ll recall, Newsom made all this particularly easy for the plaintiff, Christopher Kohls, by directly stating that he was signing the law in an attempt to force Elon Musk and ExTwitter to remove a parody video that Kohls had created.

The court cites the recent Supreme Court Moody rulings as enabling it to find that state laws targeting online speech violate the First Amendment. In response, California had argued that the law only covered unprotected “false” speech. Except, of course, it’s only a very narrow set of false speech that falls outside of First Amendment protections, and this law is hardly narrowly tailored to focus on just that speech… as Newsom himself confirmed in claiming that the parody video was “illegal” under the law.

The judge isn’t buying California’s argument at all:

While Defendants attempt to analogize AB 2839 to a restriction on defamatory statements, the statute itself does not use the word “defamation” and by its own definition, extends beyond the legal standard for defamation to include any false or materially deceptive content that is “reasonably likely” to harm the “reputation or electoral prospects of a candidate.” Cal. Elec. Code § 20012(b) (emphasis added). At face value, AB 2839 does much more than punish potential defamatory statements since the statute does not require actual harm and sanctions any digitally manipulated content that is “reasonably likely” to “harm” the amorphous “electoral prospects” of a candidate or elected official….

Moreover, all “deepfakes” or any content that “falsely appear[s] to a reasonable person to be an authentic record of the content depicted in the media” are automatically subject to civil liability because they are categorically encapsulated in the definition of “materially deceptive content” used throughout the statute. Id. § 20012(f)(8). Thus, even artificially manipulated content that does not implicate reputational harm but could arguably affect a candidate’s electoral prospects is swept under this statute and subject to civil liability.

The statute also punishes such altered content that depicts an “elections official” or “voting machine, ballot, voting site, or other property or equipment” that is “reasonably likely” to falsely “undermine confidence” in the outcome of an election contest. Id. § 20012(b)(1)(B), (D). On top of these provisions lacking any objective metric and being difficult to ascertain, there are many acts that can be “do[ne] or [words that can be] sa[id]” that could harm the “electoral prospects” of a public official or “undermine confidence” in an election…. Almost any digitally altered content, when left up to an arbitrary individual on the internet, could be considered harmful. For example, AI-generated approximate numbers on voter turnout could be considered false content that reasonably undermines confidence in the outcome of an election under this statute. On the other hand, many “harmful” depictions when shown to a variety of individuals may not ultimately influence electoral prospects or undermine confidence in an election at all. As Plaintiff persuasively points out, AB 2839 “relies on various subjective terms and awkwardly-phrased mens rea,” which has the effect of implicating vast amounts of political and constitutionally protected speech….

Thus, it’s no surprise that this law, which clearly impacts speech, must pass strict scrutiny if it’s to be found to be constitutional. This law and strict scrutiny are in different zip codes. The court finds that it is not narrowly tailored. Indeed, the Court rightly notes that any attempt to regulate political speech is particularly fraught and must be very clearly narrowly tailored, which this law was not:

The political context is one such setting that would be especially “perilous” for the government to be an arbiter of truth in. AB 2839 attempts to sterilize electoral content and would “open[] the door for the state to use its power for political ends.” Id. “Even a false statement may be deemed to make a valuable contribution to public debate, since it brings about ‘the clearer perception and livelier impression of truth, produced by its collision with error.’” Id. (quoting New York Times Co., supra, at 279, n. 19). When political speech and electoral politics are at issue, the First Amendment has almost unequivocally dictated that Courts allow speech to flourish rather than uphold the State’s attempt to suffocate it.

Upon weighing the broad categories of election related content both humorous and not that AB 2839 proscribes, the Court finds that AB 2839’s legitimate sweep pales in comparison to the substantial number of its applications, as in this case, which are plainly unconstitutional. Therefore, the Court finds that Plaintiff is likely to succeed on a First Amendment facial challenge to the statute.

Separately, the court finds that the “compelled disclosure” of manipulated images is also unconstitutional, as it is unduly burdensome compelled speech:

For parody or satire videos, AB 2839 requires a disclaimer to air for the entire duration of a video in text that is no smaller than the largest font size used in the video. Cal. Elec. Code § 20012(b). In Plaintiff Kohls’ case, this requirement renders his video almost unviewable, obstructing the entirety of the frame. Compl. ¶ 98. The obstructiveness of this requirement is concerning because parody and satire have relayed creative and important messages in American politics. As the Supreme Court has noted, “[d]espite their sometimes caustic nature, from the early cartoon portraying George Washington as an ass down to the present day, graphic depictions and satirical cartoons have played a prominent role in public and political debate.” Hustler Magazine v. Falwell, 485 U.S. 46, 54 (1988).

Defendants do not argue that Plaintiff Kohls’ video qualifies as commercial speech and the Court does not find Plaintiff’s parody to be an actual advertisement. While an argument could be made that some parodies or satire are in effect commercial speech, a vast majority of these creations are simply humorous artistic endeavors which are not subject to commercial speech regulations. In a non-commercial context like this one, AB 2839’s disclosure requirement forces parodists and satirists to “speak a particular message” that they would not otherwise speak, which constitutes compelled speech that dilutes their message.

The conclusion is clear: yes, the concern about deepfakes may be real, but you don’t get to just outlaw them under the First Amendment.

The Court acknowledges that the risks posed by artificial intelligence and deepfakes are significant, especially as civic engagement migrates online and disinformation proliferates on social media. Against this backdrop, the Court does not enjoin the state statute at issue in this motion lightly, even on a preliminary basis. However, most of AB 2839 acts as a hammer instead of a scalpel, serving as a blunt tool that hinders humorous expression and unconstitutionally stifles the free and unfettered exchange of ideas which is so vital to American democratic debate.

Just as the Court is mindful that legislative leaders enacted AB 2839 and that the State may have a legitimate interest in protecting election integrity, it is equally mindful that the First Amendment was designed to protect citizens against prior restraints and encroachments of speech by State governments themselves. “[W]hatever the challenges of applying the Constitution to ever-advancing technology, the basic principles” of the First Amendment “do not vary” and Courts must ensure that speech, especially political or electoral speech, is not censored for its ideas, subject matter, or content.

This is a good, clean, and clear ruling on a law that was shockingly unconstitutional. We can agree that deepfakes are something to be worried about. And we can even agree that Kohls’ brand of “humorous” deepfakes are not funny or compelling. But the First Amendment says that he should be free to make them, and California cannot outlaw such political speech.

I imagine that California Attorney General Rob Bonta will appeal this decision, just like he’s appealed multiple other decisions in recent years over unconstitutional First Amendment-violating laws from the California Assembly. But it sure would be nice if, rather than wasting our taxpayer money, he focused on educating both Newsom and the California legislature how the First Amendment works.

If he’d like to set up a “First Amendment 101 CLE,” I know some folks who could help.

* You may recall that the lawsuit is also challenging AB 2655. However, that law doesn’t go into effect for a while, so the immediate focus was on 2839, which went into effect immediately.

Filed Under: 1st amendment, ab 2655, ab 2839, ai, california, christopher kohls, deepfakes, gavin newsom, parody, rob bonta