Artificial Intelligence & Machine Learning , Legislation & Litigation , Next-Generation Technologies & Secure Development

US Judge Blocks California's Law Curbing Election Deepfakes

A U.S. federal judge mostly stopped from going into effect a newly-enacted California law restricting the use of election-related deepfakes, ruling Wednesday the statute likely violates American freedom of speech guarantees.

See Also: New OnDemand Webinar | Overcoming Top Data Compliance Challenges in an Era of Digital Modernization

Signed into law on Sept. 17 as part of legislative push to curb election disinformation, AB 2839 required online platforms to remove or label AI-manipulated content 120 days before the election and, in some cases, 60 days afterward (see: California Enacts Laws to Combat Election, Media Deepfakes).

U.S. District for the Eastern District of California Judge John A. Mendez granted a preliminary injunction in a lawsuit filed the same day as Democratic Gov. Gavin Newsom signed the bill. Plaintiff Chris Kohls, who goes by "Mr. Reagan" online, uses artificial intelligence to generate political videos including fake footage of Democratic presidential candidate Kamala Harris. Kohls said AI deepfakes are protected by the First Amendment, adding that the new law censored free speech and allowed people to take legal action over content they did not like.

Mendez agreed, writing the legislation "acts as a hammer instead of a scalpel," adding that it was a "blunt tool that hinders humorous expression and unconstitutionally stifles the free and unfettered exchange of ideas."

Kohls' online activity created momentum in the statehouse to pass the bill, especially after Elon Musk, owner of social media platform X - formerly Twitter - retweeted at least a Kohls videos. Newsom vowed to make manipulative electoral content illegal, writing, "I'll be signing a bill in a matter of weeks to make sure it is."

The bill language contained a safe harbor provision for satirical manipulated videos, but only so long as they contained a disclaimer visible for the video's entire duration in a font no smaller than the largest font size in the video. Mendez said that amounts to the government unconstitutionally compelling speech. Even if AI-generated content were, as California attorneys argued, subject to a lower degree of scrutiny for freedom of speech trespasses, the font requirement would still be untenable, Mendez said. The size requirement "would take up an entire screen, which is not reasonable because it almost certainly 'drowns out' the message - a parody or satire - the video is trying to convey," he wrote.

Mendez let stand one section of the bill, a requirement requiring audio-only manipulated content to contain an audible disclosure at the start of the putative recording, at the end, and at intervals of every two minutes. The disclosure may constitute compelled speech, but "a verbal disclosure at the outset and conclusion of a recording combined with interspersed disclosures in two-minute intervals is on its face reasonable and not unduly burdensome."

Newsom spokesperson Izzy Gardon told reporters the governor's office is "confident" of the court upholding the regulation, adding that the law was reasonable and similar to the one Alabama enacted. As of June, Voting Rights Lab tracked 118 bills in 42 state legislatures containing provisions intended to regulate the potential for AI to produce election disinformation.

"Deepfakes threaten the integrity of our elections, and these new laws protect our democracy while preserving free speech. Satire remains alive and well in California - even for those who miss the punchline," the spokesperson said.