31 States Enact Patchwork of Laws Regulating AI Deepfakes in Political Campaigns

The 2026 midterm elections are fast approaching, bringing a wave of high-stakes gubernatorial and congressional races across the country. For the past two years, cybersecurity experts and political analysts have warned that this cycle could be overwhelmed by artificial intelligence, predicting a flood of deceptive synthetic media designed to trick voters. But as citizens prepare to head to the polls, they are not walking into an unregulated digital wild west. Instead, they are being protected by a massive, rapid legislative response that empowers them to separate fact from fiction with confidence.[1][3]

In a rare display of bipartisan consensus and legislative speed, 31 U.S. states have enacted laws specifically regulating the use of AI deepfakes in political campaigns. This state-led initiative represents one of the fastest adoptions of technology regulation in American history. By establishing clear rules of the road, state lawmakers have transformed a looming threat into a manageable, transparent aspect of modern campaigning. This ensures voters remain capable and informed rather than anxious and deceived as they head to the polls.

The National Conference of State Legislatures, which tracks this rapidly evolving landscape, notes that states have generally adopted two distinct strategies to protect voters: mandatory disclosures and targeted prohibitions. Rather than attempting to ban artificial intelligence outright—a move that would be both technologically impossible and legally fraught—the vast majority of these 31 states have embraced transparency as the ultimate antidote to deception. This approach treats voters as capable adults who simply need accurate labeling to make up their own minds.

Disclosure laws, which represent the most common legislative approach, operate on a remarkably simple premise: campaigns are free to use generative AI, but they must explicitly tell the voter they are doing so. Generative AI refers to systems capable of autonomously creating text, images, audio, and video by analyzing patterns in massive training datasets. Under these transparency laws, if a campaign uses these powerful tools to generate a political advertisement, the ad must carry a clear, conspicuous visual or audio watermark stating that the content is synthetic.

Michigan offers a prime example of this transparency-first model in action. The state's law allows political action committees and candidates to utilize AI-generated content, provided the media includes a prominent disclosure statement. Early tests of the law during the 2026 cycle have shown highly encouraging compliance, with campaigns releasing fully AI-generated ads that clearly display the required disclaimers. This transparency allows voters to evaluate the core message of the advertisement while fully understanding its synthetic origins, neutralizing the element of trickery.[2]

Some states have taken this disclosure requirement a step further by mandating technical transparency behind the scenes. Colorado and Washington, for example, require campaigns to include metadata tags—hidden descriptive information embedded directly in the digital file's code—detailing the specific AI tools used and the exact time the synthetic media was created. This ensures that even if a visual watermark is maliciously cropped out by a third party, digital forensics can still identify the file's true origin, empowering social media platforms and journalists to verify authenticity.

The second major legislative approach involves targeted prohibitions. While disclosure focuses on broad transparency, prohibition laws aim to stop materially deceptive deepfakes—manipulated media that realistically portrays something that never actually occurred—from circulating in the critical final days of a campaign. Texas and Minnesota, for instance, explicitly prohibit the publication of deceptive political deepfakes within a specific window, typically 30 to 90 days before an election. This is the exact period when voters are most vulnerable to last-minute misinformation.

Maryland has implemented one of the most comprehensive shields in the nation, prohibiting deceptive deepfakes related to elections year-round, rather than just in the immediate run-up to voting day. These targeted bans are specifically designed to prevent so-called 'October surprises.' In these scenarios, a fabricated audio clip or video is released too close to Election Day for the victimized candidate to effectively debunk it, or for the media to fact-check it before voters cast their ballots.

Enforcement mechanisms for these laws empower those targeted by deceptive media to take immediate, decisive action. In most of the 31 regulated states, a candidate depicted in an undisclosed deepfake can seek rapid injunctive relief—a court order forcing the immediate removal of the offending advertisement from the internet or airwaves. Additionally, violators face escalating civil penalties, which in states like Michigan range from $250 for a first offense to $1,000 for subsequent violations, creating a strong financial deterrent against deceptive campaign practices.[2]

The rapid rollout of these laws has not been without growing pains, particularly regarding the First Amendment. Free speech advocates have successfully argued that overly broad deepfake bans can inadvertently criminalize protected speech, such as political satire, parody, and legitimate commentary. In California, a federal judge temporarily blocked enforcement of a broad deepfake law, ruling that it was not narrowly tailored enough to protect comedic or satirical expression. This highlights the delicate constitutional balance lawmakers must strike.

This judicial feedback loop has actually strengthened the overall regulatory framework across the country. States are actively refining their statutes to ensure they target only fraudulent voter deception while explicitly exempting satire, parody, and bona fide news reporting. This careful balancing act ensures that voters are protected from malicious manipulation without sacrificing the robust, often humorous political discourse that characterizes American elections. The ultimate result is a much more resilient, constitutionally sound set of guardrails for the public.

While states have successfully built this protective patchwork, a complex dynamic is unfolding between state capitals and the federal government. Currently, there is no comprehensive federal law regulating political deepfakes, leaving states to step up and act as the primary laboratories of democracy. However, several proposed pieces of federal legislation aim to establish a unified national standard, with proponents arguing that a single, cohesive set of rules is necessary for modern, multi-state political campaigns to function efficiently.[3]

This federal push has sparked a fierce debate over preemption. Certain federal proposals include clauses that would automatically block the application of state-level AI laws, replacing them with a single national framework. State lawmakers and advocacy groups have pushed back vigorously against these preemption clauses. They argue that federal preemption would erase the robust, tailored protections states have already successfully implemented, potentially leaving voters significantly more vulnerable if the federal standards prove to be weaker or harder to enforce.[3]

Despite these jurisdictional debates, the practical reality on the ground is highly encouraging for the 2026 electorate. Campaigns are actively adapting their workflows to comply with state disclosure requirements, conducting state-by-state audits of their media, and training staff on synthetic media identification. The normalization of AI disclaimers is effectively training the public to look for transparency labels, much like they have learned to look for traditional 'I approved this message' tags at the end of television commercials over the decades.

Furthermore, election administrators are finding that artificial intelligence is not solely a threat; it is also a powerful tool for improving the democratic process itself. Local election offices are safely utilizing large language models for low-risk, verifiable tasks such as translating voting materials into multiple languages, drafting routine voter communications, and assisting with ballot proofing. This pragmatic adoption is making election administration significantly more efficient and accessible for diverse communities, showcasing the positive potential of the technology.

As the 2026 midterms approach, the narrative surrounding artificial intelligence in politics has fundamentally shifted from one of inevitable chaos to one of managed transparency. What was once viewed as an unstoppable wave of deception has been met with a swift, effective, and bipartisan legislative wall. By prioritizing transparency and targeted accountability, these 31 states have ensured that technology serves to inform the electorate, rather than mislead it, proving that democratic institutions can successfully adapt to the AI era.[3]