The European Commission has published its final guidelines on transparency obligations for AI systems, clarifying Article 50 of the AI Act just under two weeks before the provisions become enforceable on August 2, 2026. The guidance is designed to tell providers and deployers exactly what compliance requires before the deadline arrives.
What providers must do
Under the guidelines, providers must design AI systems so that users are informed they are interacting with a machine rather than a human. They must also embed machine-readable marks in outputs so that synthetic content can be detected as AI-generated or artificially manipulated — a technical baseline meant to make provenance verifiable downstream, not merely disclosed in fine print.
What deployers must do
Obligations on deployers — the businesses and institutions that put these systems to use — are more visible to the public. They must disclose deepfakes, label AI-generated publications on matters of public interest where there has been no human editorial review, and notify individuals subjected to emotion-recognition or biometric-categorization systems. The intent is to keep people from being persuaded, profiled or deceived by AI without knowing it.
A voluntary compliance route
The Commission shipped the guidelines alongside a dedicated Q&A on Article 50 and a final Code of Practice on Transparency of AI-Generated Content. Signatories to the code can use it to demonstrate compliance without having to prove conformity by other means — a carrot that rewards firms for adopting the Commission's preferred technical approach early.
The clock is the point
Article 50 sits among the AI Act provisions phasing in on a staggered timetable, and the August 2 date has been fixed for months. Issuing detailed guidance this close to it removes a favorite industry defense — that the requirements were too vague to implement — and shifts the burden onto providers and deployers to show they are ready.