US States Push to Regulate AI Deepfakes in Elections

By early 2026, at least 26 states had enacted laws to regulate AI-generated deepfakes in political campaigns, a significant increase from just five states before 2024. These laws primarily require disclosure that content has been manipulated by AI, though states like Minnesota and Texas have opted for outright prohibitions on publishing political deepfakes within a certain timeframe before an election. The legislative push gained momentum after incidents like the deepfake robocall of President Biden in New Hampshire, which led to 26 criminal charges against the individual responsible and a $6 million FCC fine. Penalties for violating these state laws vary widely, from injunctive relief, which allows a court to order the removal of the content, to civil fines and criminal charges, including misdemeanors and felonies. Texas, a pioneer in this legislation since 2019, classifies the creation and distribution of a deepfake video to influence an election within 30 days of that election as a Class A misdemeanor, punishable by up to a year in jail and a $4,000 fine. California's approach has been particularly aggressive, with a trio of 2024 laws known as the "Defending Democracy from Deepfake Deception Act." These laws require social media platforms to label or block deceptive election-related deepfakes and hold both platforms and creators legally accountable. However, this approach has faced legal challenges; a federal judge granted a preliminary injunction against one of the California laws (AB 2655) in August 2025, citing First Amendment concerns and potential conflicts with Section 230, which protects online platforms from liability for third-party content. Washington state has also taken a broad approach, criminalizing the distribution of any malicious "forged digital likeness" intended to defraud, harass, or threaten, not just those in a political context. Effective July 2025, this law carries penalties of up to 364 days in jail and a $5,000 fine. This law supplements existing Washington statutes that mandate disclosure on deepfaked political ads and criminalize nonconsensual sexual deepfakes. Deepfake incidents have already impacted elections globally, including fabricated audio of a candidate in Slovakia discussing plans to rig an election and a manipulated clip in Nigeria falsely implicating a presidential candidate in ballot manipulation. In the U.S., beyond the Biden robocall, a deepfake was used to clone the voice of a Chicago mayoral candidate to make it seem as if he condoned police violence. At the federal level, progress has been slower. While several bills like the DEEPFAKES Accountability Act and the REAL Political Advertisements Act have been introduced to mandate disclaimers on AI-generated content, none have become law. The first federal law to address deepfakes, the TAKE IT DOWN Act, was signed in May 2025 and focuses on non-consensual intimate imagery rather than political speech. The European Union's AI Act provides a more comprehensive framework, requiring clear labeling of all AI-generated content, including deepfakes, and prohibiting manipulative uses of AI. Similarly, China has implemented rules that mandate labeling, require user identity verification to prevent misuse, and hold service providers accountable for monitoring content. This patchwork of state laws creates a complex compliance landscape for campaigns and tech platforms operating nationwide. While some laws target only the creators of deepfake ads, others extend liability to the platforms that disseminate them, creating a significant legal risk for media companies. The ongoing legal battles, like those in California and Minnesota, highlight the tension between regulating deceptive content and protecting First Amendment rights to political speech, including parody and satire.