For voice agents the obligation is direct: a person must be informed they're talking to an AI, in a clear and distinguishable manner, at the latest at the first interaction, unless it's obvious to a reasonable person. In practice, a compliant AI voice agent opens the call by identifying itself as AI.
A second limb covers synthetic audio: deployers generating realistic AI voices must disclose that the audio is artificially generated. Penalties for breaching the transparency rules reach up to €15M or 3% of worldwide annual turnover.
The practical takeaway is that disclosure is a built-in feature, not a constraint. Tests consistently show that telling the customer it's AI has no measurable impact on conversion, what matters is being helpful, not pretending to be human. Callsy discloses on every call by default.