EU AI Act Article 50

What it is

Article 50 of the European Union's AI Act sets transparency duties that do not depend on an AI system being high-risk. As the text reads, providers must design systems that interact directly with people so that those people are informed they are interacting with an AI, unless that is obvious to a reasonably well-informed person given the circumstances. Providers of systems that generate synthetic audio, image, video or text must mark the output as artificially generated in a machine-readable way. Deployers have further duties for deepfakes and for AI-generated text on matters of public interest.

For a voice agent calling or answering in the EU, the practical reading is: say it is an AI at the start of the call, and be able to show how the synthetic audio is marked. The obligations apply from 2 August 2026. Proposed "Digital Omnibus" amendments may change timing or scope, so this row is flagged as subject to change.

Why it matters when buying

Disclosure is now a design requirement rather than a courtesy, and it interacts with turn-taking: a caller who interrupts the disclosure must still have been told. Vendors serving the EU should have a default disclosure line per language and a record that it played. The Act's separate high-risk rules may also apply to agents used for credit decisions, essential-service eligibility or recruitment. Informational only; confirm with counsel and the regulator.

What to ask

Ask for the default disclosure wording in each language you serve and whether it can be interrupted. Ask how the platform records that disclosure was delivered on each call. Ask how synthetic audio is marked and how that is evidenced. Ask whether the vendor has assessed your use case against the high-risk categories.