With Congress stalled, state legislators have taken real action on AI deepfakes.
Oct. 22, 2024 — As more than a dozen bills aimed at limiting or regulating AI-generated deepfake images languish in Congress, state legislators have stepped up to pass 57 new deepfake-related laws in 28 states.
That data, tallied by the National Conference of State Legislators, includes bills passed in both the 2023 and 2024 legislative sessions. And there may be more. Nearly every state legislature has concluded its business for the year, with the exception of Pennsylvania, Ohio, New Jersey, and Michigan.
Concern over harmful AI-generated images has been growing since the term “deepfake” first appeared in 2018. That concern grew into outrage in 2022 when new generative AI tools allowed virtually anyone to create deceptive images of real people. Harmful incidents quickly made headlines, with middle school and high school students circulating fake AI-generated non-consensual intimate images of their classmates.
Members of Congress have introduced more than 120 bills dealing with artificial intelligence, and a handful of proposals intended to rein in deepfakes have made it to the committee hearing stage. But none have passed.
Meanwhile, local legislators in 28 states passed 57 new laws aimed at protecting individuals from the cruel and harmful effects of deepfake images.
two aims: harmful imagery, and election fakes
Those state-level laws generally fall into two categories: Prohibiting the intentional dissemination of deepfake recordings or images of an intimate nature without the consent of the person depicted, and regulating the use of AI-generated or enhanced imagery related to political elections.
Some of the new laws prohibit the distribution of both genuine imagery and AI-generated or AI-enhanced imagery, as in Delaware’s H 353, enacted in Oct. 2024. The critical elements are that intimate imagery or sexual conduct is involved; and that the person depicted did not consent to disclosure of the imagery.
The election-specific laws often prohibit the dissemination of deepfake imagery depicting a political candidate within a 90-day window prior to an election. One example: Arizona’s S 1359, enacted in late May 2024.
court challenges expected
The new election-focused laws may face court challenges in the coming months, as courts have historically carved out significant First Amendment space for political speech, including the satirical portrayal of an opponent in political ads. (See, for instance, this classic 1980 Republican poke at Democratic House Speaker Tip O’Neill.)
There is a difference between past satire and AI deepfakes, though: A reasonable person in 1980 knew that wasn’t really Tip O’Neill behind the wheel. Today an AI-generated deepfake could depict a political opponent carrying out a scurrilous act—and it would look absolutely real to most viewers.
Two of California’s new deepfake laws were in fact challenged in court last month, just weeks after they were signed into law by Gov. Gavin Newsom. The early rulings in that case—and in further challenges expected in other states—may help shape the next iteration of deepfake laws expected to be introduced at the state level in early 2025.