Chatbot disclosure#A50-1-CHATBOT-DISCLOSURE
If people interact directly with an AI system (chatbot, voice agent), they must be informed they are dealing with AI — unless this is obvious to a reasonably well-informed person from the context.
Sources & methodology
A50Check assessments are generated against an encoded rulebook of the EU AI Act's Article 50 transparency obligations. Every requirement below cites the official passage it relies on — verbatim, with a deep link into the EUR-Lex consolidated text of Regulation (EU) 2024/1689 (AI Act). Paid deliverables are additionally human-reviewed before they are issued.
Rulebook encoding 0.1-draft — pending legal validation. General information, not legal advice.
Machine readable marking#A50-2-MACHINE-READABLE-MARKING
Providers of AI systems that generate synthetic audio, image, video or text must ensure outputs are marked in a machine-readable format and detectable as artificially generated or manipulated (e.g. watermarking, metadata, provenance standards), so far as technically feasible.
Emotion biometric notice#A50-3-EMOTION-BIOMETRIC-NOTICE
Organisations using emotion recognition or biometric categorisation systems must inform the people exposed to them, and process personal data in line with EU data protection law.
Deepfake label#A50-4-DEEPFAKE-LABEL
AI-generated or manipulated images, audio or video that depict real or plausibly real persons, places, objects or events — and could misleadingly appear authentic — must be visibly disclosed as artificially generated or manipulated.
Public interest text#A50-4-PUBLIC-INTEREST-TEXT
AI-generated or manipulated text published to inform the public on matters of public interest must be disclosed as artificially generated — unless the content has undergone human review or editorial control and a natural or legal person holds editorial responsibility for it.
Disclosure quality#A50-5-DISCLOSURE-QUALITY
All Article 50 disclosures must be provided in a clear and distinguishable manner, at the latest at the time of first interaction or exposure, and meet applicable accessibility requirements.
Scope non retroactive#A50-SCOPE-NON-RETROACTIVE
Content published before 2026-08-02 does not need retrospective labelling. However, republishing, substantially modifying, or reusing archived AI content in a new campaign or commercial context after that date triggers the labelling obligation afresh.
Org content map#A50-ORG-CONTENT-MAP
Organisations must know where AI generates or alters anything they publish — website imagery, social posts, chatbots, marketing copy, reports — and determine which obligations attach to each, including whether they act as provider, deployer, or both.