California's latest law could land social media users who post or repost AI deepfakes that mislead voters about upcoming elections in legal trouble. Gov. Gavin Newsom has suggested that AB 2839, which he signed into law Tuesday and took immediate effect, could be used to curb retweets that spread false content, including from Elon Musk.
“Just signed a bill making this illegal in California,” Newsom said in a tweet, a reference to an AI deepfake Musk reposted earlier this year that appeared to show Kamala Harris calling herself an incompetent candidate and diversity hire (which she's not).
“You can no longer knowingly distribute ads or other election-related communications that contain materially deceptive content, including deepfakes,” Newsom said in a later tweet.
California's new law targets distributors of AI deepfakes, especially when the posts resemble candidates on California's ballot and the poster knows they are confusing and false. AB 2839 is unique in that it goes after those who maliciously spread AI deepfakes, not the creators of them or the platforms on which they appear. Anyone who sees an AI deepfake on social media can now file for an injunction, meaning a judge can order the poster to remove it or award monetary damages to the poster.
This is one of the strongest US laws banning election-related AI deepfakes ahead of the 2024 presidential election.
The California Initiative on Technology and Democracy (CITED), which helped draft AB 2839, told TechCrunch that the law could affect not just Musk, but any social media user who posts or reposts election-related AI deepfakes with malicious intent, meaning the poster knew it was false and would confuse voters.
“[AB 2839] “This bill would go after creators or distributors of content if that content falls under the conditions of the bill,” Leora Gershenzon, policy director at CITED, said in an interview with TechCrunch. “This is materially deceptive content that is knowingly false and distributed with a reckless disregard for the truth that could potentially influence an election.”
Asked whether Musk could face legal action for reposting the deepfake, Newsom did not rule out the possibility.
“I think Musk missed the punchline,” Newsom said at a press conference on Thursday. “Parody is alive and well in California, but deepfakes and election manipulation undermine our democracy.”
Specifically, the new law would ban election-related AI deepfakes on television, radio, phone, text, or any communication “delivered via the internet.” The bill is not limited to political campaign ads, as other laws have focused on, but also covers posts by members of the public. AB 2839 establishes a period from 120 days before a California election to 60 days after the election during which stricter rules apply to what can and can't be posted on social media about political candidates.
“The real goal isn't damages or an injunction,” Gershenson said, “but to stop people from doing it in the first place. That would actually be the best outcome, to stop deepfakes from tampering with elections.”
The law applies to state and local candidates in California, as well as federal candidates such as Kamala Harris and Donald Trump who appear on California's ballot. AB 2839 does not apply if the AI deepfake contains a clear disclaimer indicating it has been digitally altered.
Musk has already tested California's willingness to enforce the new law by reposting a deepfake likeness of Kamala Harris, whom Newsom mentioned in a tweet, on Tuesday, garnering more than 31 million impressions on X. Musk also reposted an AI deepfake likeness of Newsom on Wednesday, garnering more than 7 million impressions.
Musk and X have also faced other legal issues related to moderation. For example, a Brazilian Supreme Court judge on Thursday fined X Corp for evading the country's ban on the platform. The judge had previously said that X's failure to tackle fake news and hate speech was undermining Brazilian democracy.