AI Act Transparency Rules (Article 50): Chatbots, Deepfakes and AI-Generated Content

Four duties, two roles

The information must be given in a clear and distinguishable manner, at the latest at the first interaction or exposure, and meet accessibility requirements (Article 50(5)). The duties apply alongside the high-risk rules and other transparency rules in EU or national law (Article 50(6)).

Provider or deployer? if you build a customer chatbot on a model API and offer it under your name, you are usually the provider of that AI system, so Article 50(1) is yours to meet. If you use a vendor's tool, you are the deployer, and your duties are 50(3) and (4). See the AI Act timeline for the role test.

Dates

The exceptions

• Obvious AI: no interaction notice is needed where it is obvious from the circumstances and context (50(1)).
• Assistive editing: the marking duty does not apply to the extent a system performs an assistive function for standard editing or does not substantially alter the input data or its meaning (50(2)).
• Creative works: for evidently artistic, creative, satirical or fictional deepfakes, disclosure is limited to the existence of generated content, in a way that does not hamper the work (50(4)).
• Edited AI text: the disclosure for AI text on public-interest matters does not apply where the text has undergone human review or editorial control and someone holds editorial responsibility for the publication (50(4)).
• Law enforcement: systems authorised by law to detect, prevent, investigate or prosecute criminal offences are largely exempt, with conditions (50(1) to (4)).

Code of practice and guidelines

Article 50(7) asks the Commission to encourage codes of practice on detecting, marking and labelling AI-generated content. A Code of Practice on Transparency of AI-Generated Content was published in June 2026; in July 2026 the Commission, with the AI Board, assessed it as adequate, and published its final guidelines on Article 50. Following the code is a practical way to implement the marking and labelling duties, but the Digital Omnibus notes that codes of practice have limited legal effect and do not grant a presumption of conformity (recital 41 of Regulation (EU) 2026/1744). Authorities still look at what you actually do.

Enforcement

Infringements of Article 50 by providers and deployers carry fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher (Article 99(4)(g)); for SMEs and start-ups, whichever is lower (Article 99(6)), and the Omnibus extends that to small mid-cap enterprises (Article 99(6a)).

Checklist

1. List every AI system you offer or use that interacts with people, generates content, or recognises emotions or categorises people biometrically, and note your role for each.
2. Chatbots and voice agents you provide: show a short notice at the start of every conversation, such as "You are chatting with an AI assistant", and make it accessible.
3. Generative features you provide: add machine-readable marking (for example metadata or watermarking) to output; for systems on the market before 2 August 2026, by 2 December 2026.
4. Marketing and communications: label deepfake images, audio and video; label AI-written public-interest articles unless they are edited under someone's editorial responsibility.
5. Emotion recognition or biometric categorisation: inform people before exposure and check the GDPR basis; emotion recognition at work and in education is prohibited outright except for medical or safety reasons (Article 5(1)(f)).
6. Record the decisions in your AI register and your AI literacy material, so staff know when a notice is needed.

Transparency rules, written down. The AI Agent Governance Pack includes an agent use policy with AI literacy and transparency rules, a register of AI agents with each one's AI Act role, a DPIA template for agentic processing and an incident playbook.

Sources

• Regulation (EU) 2024/1689 (AI Act), Articles 5, 50, 99, 111 and 113, EUR-Lex
• Regulation (EU) 2026/1744 (Digital Omnibus on AI), Article 1 (amendments to Articles 50(7), 99 and 111) and recitals, EUR-Lex
• Faegre Drinker: Commission confirms transparency code of practice as adequate and publishes final Article 50 guidelines (July 2026)

Related tools