What's required today (federal)
- Consent: AI voices are artificial voices (FCC 24-17), so calls need prior express consent, and prior express written consent for marketing.
- Identification: every artificial-voice message must state the business's identity at the start and give a callback number (47 CFR 64.1200(b)).
- Opt-out: marketing messages must offer an automated opt-out.
- No specific "I'm an AI" rule: despite what some vendors and complaints say, no final federal rule requires the call to announce that it's AI.
What's pending: FCC 24-84
In August 2024 the FCC proposed rules (FCC 24-84) that would:
- define an "AI-generated call";
- require disclosure, when consent is collected, that it may cover AI-generated calls or texts;
- require a disclosure at the start of each AI-generated call; and
- exempt people with speech or hearing disabilities who use AI to place their own calls (not for telemarketing).
Comments closed in late 2024. As of October 2026 there's no final rule. Building for disclosure now is a low-cost hedge.
State rules
- Utah (AI Policy Act, as amended in 2025): generative AI must disclose it's AI when a consumer asks, and regulated occupations must disclose proactively.
- California: bots used to incentivize a sale must disclose they're bots in certain online communications. Separately, California's privacy law (CIPA) creates recording and eavesdropping risk for AI voice vendors.
- Colorado rewrote its AI Act in 2026 (effective 2027). Check the final text for any interaction-disclosure duty.
State AI laws change quickly. Verify against the state legislature's site before relying on them.
FAQ
Does the FCC require AI callers to say they're AI?
No final rule requires it yet. The FCC proposed an AI-call disclosure requirement in 2024 (FCC 24-84), which is still pending. AI-voice calls must already identify the business and have proper consent.
Do I need to tell people my AI agent is AI if they ask?
In some states, yes. Utah requires generative AI to disclose it's AI when a consumer asks. Honesty is also the safer course under consumer-protection law generally.
Is there a way to avoid the disclosure question entirely?
Have a person make the call. If your AI agent writes the brief and a human dials and speaks, there's no AI voice on the call.
Sources
- FCC 24-84 NPRM (AI-generated calls) · FCC 24-17
- 47 CFR 64.1200(b)
- FCC proposes AI call rules (Kelley Drye)
General information, not legal advice.
Your agent briefs, a person calls.