Twelve governors from both parties signed a compact Wednesday criminalizing deceptive synthetic audio in campaign communications, the first coordinated state-level response to AI-generated voice clips mimicking candidates. The agreement sets disclosure standards, 48-hour platform takedown requirements, and felony penalties when fabricated audio targets election administration or voter turnout.
Signatories include battleground-state executives from Michigan, Pennsylvania, Wisconsin, Arizona, and Nevada alongside coastal states that already regulate deepfake pornography. Republican Governor Brian Kemp of Georgia and Democratic Governor Gretchen Whitmer of Michigan co-chaired negotiations after separate incidents in 2025 and early 2026 in which robocalls using synthetic voices instructed voters to stay home on election day.
Compact Terms
Campaign committees must label any audio or video containing generative AI with a standardized on-screen and spoken disclosure within the first three seconds. Platforms operating in signatory states must remove reported deceptive clips within 48 hours when a bipartisan state board certifies the filing is authentic.
Intent matters for criminal liability. Parody and satire with clear labels remain protected. Prosecutors must show creators designed content to deceive listeners about a candidate's statements or official election procedures. Penalties reach five years for repeat offenses tied to voter suppression.
Why States Moved First
Congress has not passed federal election deepfake legislation despite bipartisan draft bills in both chambers. The FEC deadlocked on whether existing finance rules cover generative content costs. State officials said they could not wait for a polarized Washington while cheap voice-cloning tools proliferated.
Secretaries of state reported a sixfold increase in AI-related complaints between January and May 2026 compared with the entire 2024 cycle. Most incidents involved audio rather than video because voice models require less compute and spread faster through messaging apps.
Platform and Legal Challenges
Meta and Google policy teams participated as observers but did not sign the compact. Company spokespeople said uniform national rules remain preferable to state-by-state takedown clocks. First Amendment scholars noted felony thresholds may face court challenges when content lacks explicit calls to abstain from voting.
The ACLU warned that broad criminal statutes could chill legitimate political satire. Compact authors inserted a safe-harbor provision requiring prosecutors to certify that no reasonable viewer would interpret labeled satire as factual before bringing charges.
Partisan Dimensions
Republican signatories emphasized fraud prevention and ballot security. Democratic signatories stressed voter intimidation and minority-community targeting. Both frames produced identical technical requirements, a rarity in recent election-law debates.
Non-signatory states include Texas and Florida, where leaders argued existing impersonation statutes suffice. Compact supporters said offenders will route content through non-signatory jurisdictions, pushing the group to lobby Congress for federal baseline standards.
Impact on 2026 Midterms
Down-ballot races face the first widespread test of enforcement machinery before the presidential cycle. State boards will publish public logs of takedown requests to deter bad actors. Campaign consultants said disclosure rules add production steps but reduce reputational risk from undisclosed synthetic ads.
Federal candidates operating across signatory and non-signatory states must navigate multiple compliance regimes or adopt the strictest standard nationwide. Political ad agencies expect a premium for verified human-voice recordings and blockchain timestamping services marketed as fraud insurance.
