AI disclosure: definition for voice AI

Definition

AI disclosure is telling a person that they are talking to an automated system, not a human. No single law sets one rule for every call. The EU AI Act, FCC rules for artificial voices and state laws each address it differently, and one FCC rule is still only proposed.

What it covers

AI disclosure is a duty, where one exists, to inform a person that they are interacting with an AI system. It is not the same as consent to be called or recorded, and it is not the same as the caller identification that US rules already require for artificial voice messages. Different instruments cover different parts, so the useful question is which one reaches a given call.

What it requires

European Union. Article 50(1) of the AI Act requires providers to design AI systems that interact directly with people so that those people are informed they are interacting with an AI system, unless that is obvious to a reasonably well-informed, observant and circumspect person given the context. Article 50(5) says the information must be clear and distinguishable and given at the latest at the first interaction. The Act applies generally from 2 August 2026. Regulation (EU) 2026/1744 later amended it, changed Article 50(7) on codes of practice, and gave providers of generative systems already on the market until 2 December 2026 to comply with the Article 50(2) marking duty. The amending regulation leaves the 2 August 2026 date for Article 50(1) in place. Whether a company counts as a provider, a deployer or both depends on the Act’s definitions.

United States, federal. The FCC’s 2024 ruling (FCC 24-17) treats AI-generated voices as artificial voices under the TCPA. The rules for artificial voice messages require the message to state the identity of the responsible entity at the start. That is identification, and it is not an AI disclosure. In August 2024 the FCC proposed rules (FCC 24-84) that would define AI-generated calls and require callers to disclose them. That is a proposal. As of the review date the eCFR text of 47 CFR 64.1200 has no paragraph requiring AI disclosure, and no final rule adopting the proposal has been issued in that docket.

United States, states. Some states have their own bot or AI disclosure laws. Their scope differs: some reach only online interactions or particular sectors, so read the statute’s scope, not its title, before relying on one for phone calls.

What it means for a voice agent

Most teams put the disclosure in the first sentence of the call, and write the agent’s instructions so that it answers truthfully if asked whether it is a person. That choice is a design decision as much as a legal one, and it is cheap to make early. Pair it with the spoken notice for call recording consent where one is needed.

Disclosure does not replace consent. An outbound call by an AI phone agent still needs the consent its call type requires, which is the subject of TCPA and outbound calling. If callers in the EU are reached, check the Article 50 duties for both your own role and your vendors’ roles. For the inbound side, see the guide on the AI receptionist.

Questions to ask counsel

Which disclosure rules apply to calls in each place we operate, today and on the next effective date?
Are we a provider, a deployer or both under the EU AI Act for this agent?
Is a disclosure at the start of a call enough, or must it be repeated or recorded?
Which state laws reach telephone calls, and what do their scopes say?
How should an agent answer when asked whether it is a human?

Guides

AI receptionist: what to buy, what to build

Use cases

Sources

This page is general information, not legal advice. Rules differ by country, state and use case, and they change. Check the linked sources and ask qualified counsel before you rely on it.

Last reviewed October 19, 2026