In January 2024, thousands of New Hampshire voters got a robocall in what sounded like President Biden’s voice, telling them not to vote in the state’s primary. The voice was generated by AI. The case became the first major test of how existing law handles synthetic media in elections, and it shaped much of the regulation that followed.
What happened next
- The FCC acted on robocalls. In February 2024, the Federal Communications Commission ruled that AI-generated voices count as “artificial” voices under the Telephone Consumer Protection Act, which already restricted robocalls using artificial or prerecorded voices. That made AI-voice robocalls without consent illegal under existing law.
- Enforcement followed. The political consultant behind the New Hampshire calls faced a multimillion-dollar FCC fine and state criminal charges, and the phone company that carried the calls settled with the FCC.
- States passed laws. Dozens of states now regulate AI-generated content in election communications, most commonly by requiring a disclosure label on synthetic media within a set period before an election, with exceptions for satire and news coverage. The details vary widely, and some laws have been challenged on free speech grounds.
What is still allowed
There is no general federal ban on political deepfakes. Satire, parody and clearly labeled synthetic content are generally protected speech. Many platforms require labels on AI-generated political ads, and the major AI image and video tools have policies against generating realistic images of real candidates, but enforcement is uneven.
The “liar’s dividend”
The bigger long-term risk may be the reverse problem: once everyone knows fakes are possible, real recordings can be dismissed as AI. Researchers call this the liar’s dividend. It makes verification of authentic material just as important as detecting fakes.
How to check a suspicious clip or image
- Find the first source. Who posted it first, and when? Clips with no clear origin deserve the most caution.
- Look for corroboration. A real statement by a candidate will usually be covered by multiple outlets or appear in official video.
- Check for content credentials. Some cameras, editing apps and AI tools now attach C2PA “content credentials” describing how a file was made. Inspection tools can read them, though their absence proves nothing.
- Run a reverse image search on images and key video frames to find earlier versions.
- Be skeptical of detection tools. AI detectors give false positives and false negatives. Treat them as one signal, not a verdict.
- Notice the timing. A shocking clip that appears days before a vote, when there is little time to debunk it, fits the pattern of past disinformation campaigns.
For the text side of the problem, see astroturfing and paid commenters, and for a quick worksheet, use the claim checklist.