Article 50 of the EU AI Act has applied since 2 August 2026. It requires providers of generative AI to mark outputs in a machine-readable way, and organisations that use AI to disclose deepfakes and AI-generated text published on matters of public interest.
Generative AI systems placed on the market before 2 August 2026 have until 2 December 2026 to meet the marking requirement, which makes December the date to plan around.
Key Facts
- Article 50 of the AI Act applies from 2 August 2026.
- Systems placed on the market before 2 August 2026 have until 2 December 2026 to meet the Article 50(2) marking and detection obligation.
- People must be told when they are interacting with an AI system, unless it is obvious.
- Deepfakes, and AI-generated text published to inform the public on matters of public interest, must be disclosed, unless the text has had human review or editorial control.
- Fines for breaching these obligations can reach EUR 15 million or 3% of worldwide turnover.
Who has to do what under Article 50?
Duties depend on your role. Providers, the organisations that develop or place an AI system on the market, must design chatbots so users know they are talking to AI, and must mark generated audio, images, video and text so they can be detected as AI-made.
Deployers, the organisations using AI in their own work, must inform people exposed to emotion recognition or biometric categorisation, disclose deepfakes, and label AI-generated text published on matters of public interest. Many companies are both, for different tools.
What counts as labelling in practice?
For providers, labelling means technical marking, such as watermarks or metadata, that machines can read. For deployers, it means a clear, visible disclosure at the latest when people first see the content. The EU’s voluntary Code of Practice on transparency of AI-generated content describes accepted ways to do both.
How do you get ready before December?
- List every tool that generates content, including AI features built into SaaS products you already use.
- Decide your role for each tool: provider, deployer or both.
- Ask suppliers how their outputs are marked and request written confirmation.
- Set an editorial policy: which content gets human review, who signs it off, and where labels appear.
- Keep records of these decisions so you can show them to a regulator.
Our AI Act compliance and AI governance services cover this mapping, and AI policy development turns it into rules your teams can follow.
Frequently asked questions
Do we have to label content created before August 2026?
No retroactive labelling is required, although the European Commission encourages it.
Does AI-assisted marketing copy need a label?
The text duty covers AI-generated text published to inform the public on matters of public interest. It does not apply where the text has had human review or editorial control and someone holds editorial responsibility.
Is the Code of Practice mandatory?
No. It is voluntary, but signatories can rely on it to show compliance. Others must show compliance through other adequate means.
What to do next
Transparency rules are easier to meet when they are part of how you build and publish content, not a last-minute sticker. If you want a clear view of where AI is used across your business and what each use requires, contact us.
