State AI-in-elections laws every campaign should know

Beyond the FCC and TCPA, states are writing their own AI-in-elections rules. Here is the layer campaigns miss — voice-cloning bans, deepfake laws, and disclosure requirements.

Most campaigns that look at AI voice stop at the federal layer — the TCPA and the FCC. That is necessary but not sufficient. States are writing their own AI-in-elections rules, and they reach conduct the federal framework does not, especially around synthetic voices and video. If you operate in multiple states, you are operating under multiple rulebooks at once.

Voice cloning is its own category of risk

The clearest state line is on cloning a real person’s voice. Tennessee’s ELVIS Act criminalizes unauthorized use of a person’s voice, and it is not aimed only at music — a cloned candidate voice is squarely in scope. This is separate from the TCPA: even a call that had consent and disclosure can violate a voice-cloning law if it impersonates a real person.

So “we disclosed it was AI” does not cure a cloned voice. Disclosure answers the artificial-voice question; it does not grant permission to imitate a specific human. The only safe rule is to never clone a candidate’s or any real person’s voice at all.

Deepfake-in-elections laws target synthetic media

A growing set of state laws targets AI-generated deepfakes in elections — synthetic audio or video of a candidate, often with disclosure mandates or outright bans within a window before an election. These sit on top of, not instead of, the FCC’s position that AI voices are “artificial voice” under the TCPA — a stance the FCC took after a political deepfake robocall.

• Voice-cloning bans (e.g. Tennessee’s ELVIS Act) — impersonating a real voice
• Deepfake-in-elections laws — synthetic candidate audio or video
• State disclosure mandates — labeling AI-generated communications
• State calling-time windows and do-not-call rules, layered on federal ones

Build the discipline in, don’t bolt it on

Because the rules vary by state and keep changing, the durable answer is a tool that keeps consent and disclosure as a matter of design, not something you remember to switch on. Teltura discloses it’s AI on every call, never clones a candidate’s voice, and keeps a compliance ledger tracking consent, disclosure, recording, and cross-channel contact frequency — a per-call record you can produce if a state regulator or an opponent asks.

None of this is legal advice, and a multi-state campaign should confirm the specifics with counsel. But the defensible posture is consistent across every state: consented lists and inbound, disclosed AI, honored windows, and no synthetic voices of real people.

The federal layer isn’t the whole map — state voice-cloning and deepfake laws add rules the TCPA never touches, and the safe posture in all of them is disclosed AI on consented lists, with no cloned voices ever.