A California judge opens the door to election deepfakes

Politico

2 days ago

American elections are increasingly clouded with anxiety over what AI-powered misinformation could be doing to democracy — misleading voters, smearing candidates, polluting the media ecosystem.
Despite Congress' inaction on the issue — or maybe because of it — more than half of the U.S.' state legislatures have either considered or passed laws on election deepfakes, making states the main guardians of what Americans can and can't circulate about elections.
A federal court, however, struck down one of the country's strictest state laws on AI election misinformation on Tuesday — raising the question of how, or whether, regulators should address deepfakes that could skew the vote.
The law, signed by Gov. Gavin Newsom last year, prohibits platforms from displaying election-related deepfakes within about four months of voting day in California. (States like Texas have similar laws, though they're mostly aimed at the campaigns themselves.)
The challenge had its roots in a clash of egos. When X owner Elon Musk shared a distorted clip of then-Vice President Kamala Harris on the site, Newsom rebuked him. A few months later, Newsom signed the law, and the video's creator promptly filed a suit against California to get the deepfake rule overturned. Musk's X later joined the suit, along with Rumble and the Babylon Bee.
On Tuesday, California District Judge John Mendez overturned the law — not on First Amendment grounds, but because it violated Section 230 of the federal Communications Decency Act, which protects platforms from liability for content that third parties post. He ruled that when a user uploaded the deepfake clip of Harris saying she was the 'ultimate diversity hire,' platforms didn't have a duty to block it.
'This is not surprising in the least,' Ari Cohn, lead tech counsel for the Foundation for Individual Rights and Expression, said of the decision. 'It's kind of bizarre to me that [California] thought they could sneak this by Section 230 in the first place.'
The full transcript of Medez's ruling likely won't be available until a month from now. But when he delivered his opinion from the bench, Mendez didn't invoke the First Amendment arguments that the attorneys raised about the law. (My colleague Chase DiFeliciantonio, the POLITICO reporter in the courtroom, told DFD that neither the websites nor the state pressed the issue further when Mendez asked.)
You might see why the ruling makes sense from a tech policy perspective — regardless of how much it might horrify election-reform advocates.
Section 230 is widely regarded as the '26 words that created the internet.' Written to protect service providers like America Online from being sued for the content their users were posting, it has evolved into the primary liability shield for massive online platforms like X, Facebook, YouTube and Reddit, which would likely be sued into oblivion if they could be liable for every single post. Without it, they might pre-emptively take down everything that has even a remote chance of being illegal, making feeds boringly vanilla. Section 230 also shields platforms when removing content they deem to be objectionable, which is essential for moderation.
At the same time, Section 230 has become a political punching bag. Some advocates say it has hugely exceeded its original intent, and now protects companies from harms they knowingly allow to fester.
The debate is especially sharp around elections, and not just from Democrats worried about Musk's provocations. During the 2020 presidential race, conservatives were up in arms when Facebook and Twitter (now X) suppressed posts about explicit videos on Hunter Biden's misplaced laptop — the kind of content policy protected by Section 230.
Now with the rise of deepfakes, that argument has broadened — and critics suggest that the law allows for a genuine degradation of democracy.
Section 230 isn't ironclad. The law doesn't absolutely shield platforms that host child sexual abuse material or copyright-infringing works, or that facilitate sex trafficking. First lady Melania Trump recently pushed through the Take It Down Act, a law signed in May that imposes criminal liability on platforms that display sexual deepfakes and revenge porn. (Section 230 technically doesn't apply to criminal laws.)
So why shouldn't there be an exception for election deepfakes, too?
For one, the country has a long tradition of trying to promote, and protect, robust political discourse. Pornography, by contrast, has generally been considered as less worthy of protection. 'Congress' judgment is if the consequences are that platforms will over-restrict porn in some measure, no big deal,' UCLA Law professor Eugene Volokh, a First Amendment specialist, told DFD.
Whereas when it comes to political speech, even AI-generated, he says: 'We might be worried about the over-chilling of, say, election-related parody.'
There's even a counterintuitive argument that such an exception could actually make misinformation worse. Cohn points out that without Section 230, a politician could file a bunch of lawsuits against platforms with frivolous claims that an unflattering video — even a perfectly real one — was a deepfake. In response to a 2024 election ad that compiled clips of then-candidate Donald Trump's real-life gaffes, he falsely asserted that they had been fabricated using AI.
There may be ways to construct a narrower law than California's that would pass muster. A law that restricts election deepfakes a week, rather than four months, before the vote may be more legally defensible, according to Cohn. Courts have also left open the possibility that restrictions on misinformation about election mechanics — such as misleading people about how to vote — could be permissible. If there were a deepfake video of Biden telling people to visit the polls on the wrong day, served up to likely Democratic voters, a law could conceivably be narrow enough to address that kind of attempt to disrupt the system.
For legislators looking to solve the deepfake problem, there's one other nettlesome fact: Despite all the anxiety, there isn't much evidence that election deepfakes in fact skewed the 2024 vote. Without having a documented harm, it's tough to know how to create a focused, defensible law to regulate it.
Cohn said, 'The price of living in a free society is that bad things are sometimes going to happen before we can fix them.'
AWS gives the administration a $1 billion coupon
Amazon Web Services is effectively chipping in $1 billion to advance Trump's AI agenda, POLITICO's Sophia Cai reports.
The General Services Administration announced on Thursday that it had struck an agreement with AWS to help move federal agencies onto the cloud. The deal, which is set to run through 2028, provides for $1 billion in credits for AWS' services. More access to cloud computing will help agencies to integrate large language models and other deep learning applications into its operations.
The coupon is part of DOGE's OneGov initiative, which seeks to digitize the federal government's paper-based systems. Oracle also agreed to give agencies a 75 percent discount on its cloud and database services in July, and a senior administration official told POLITICO that similar OneGov deals are in the works with Google and Microsoft.
Trump announces a 100 percent chip tariff
Trump says he'll impose 'a tariff of approximately 100 percent' on all semiconductors manufactured abroad, POLITICO's Doug Palmer reports.
During a Wednesday press conference with Apple CEO Tim Cook, the president indicated that the import tax was intended to pressure more semiconductor firms to come to the U.S. 'If you're building in the United States of America, there's no charge,' Trump said. He added, 'The chip companies are all coming back home.'
A White House official told POLITICO that the tariffs will be designed to reshore manufacturing while also limiting disruptions to the supply chain.
Trump has previously used tariffs as leverage to increase investment in domestic chip manufacturing. He claimed in April that he'd convinced Taiwan Semiconductor Manufacturing Company to pitch in another $100 billion for its expansion into the U.S. by threatening the company with a 100 percent tariff as well. The Commerce Department has also been renegotiating grants provided under the CHIPS Act to press semiconductor companies to up their investments in American manufacturing.
post of the day
THE FUTURE IN 5 LINKS
Stay in touch with the whole team: Aaron Mak ([e-mail]); Mohar Chatterjee ([e-mail]); Steve Heuser ([e-mail]); Nate Robson ([e-mail]); and Daniella Cheslow ([e-mail]).