Does Article 50 apply to your AI?
Chatbots, AI-generated content and deep fakes have had transparency duties under the EU AI Act since August — whatever their risk class. Answer up to seven questions to see exactly which ones apply to you, and how long you have left.
- Your duties, paragraph by paragraph
- Provider vs deployer, worked out for you
- Free · No signup · About 1 minute
Who built the AI system you're checking?
Putting your own name on a product built on someone else's model counts as building it.
Do people chat or talk with it, or does it act on someone's behalf?
Chatbots, voice assistants, support agents and AI agents that email, call or book on someone's behalf.
Does it generate images, audio, video or text?
Any generated output counts, including chatbot replies, summaries, product images or voice.
Could any of that content pass as real?
Realistic images, audio or video of real-looking people, places or events — a “deep fake” under the Act.
Do you publish AI-generated text to inform the public?
News, health, financial, policy or similar public-interest topics — without human editorial review before publication.
Does it recognise emotions or sort people by biometric traits?
For example, inferring mood from a voice, or age or gender from a face.
When did it go live in the EU?
Systems already on the market before 2 August 2026 have until 2 December 2026 to mark their output.
Built on someone else's model? For Article 50 you're still the provider of your system, so these duties are yours, not the model vendor's.
Breaching Article 50 can cost up to €15 million or 3% of worldwide turnover, whichever is higher (lower caps apply to SMEs).
Classify the system, keep the evidence in one place, and track every obligation. No card needed.
New to Article 50? Read the transparency guide, or check the rest of the Act with the risk quiz.
Article 50 in brief
Does Article 50 apply if my system isn't high-risk?
Yes. Article 50 isn't a risk tier. It applies to chatbots, AI content generators, emotion recognition and deep fakes whatever their risk class, and on top of the high-risk rules when a system is high-risk too.
When does Article 50 apply?
Since 2 August 2026. Generative AI systems already on the market before that date have until 2 December 2026 to mark their output in a machine-readable way (Article 50(2)).
We use OpenAI or another vendor's model. Isn't this their problem?
Usually not entirely. If you offer an AI system under your own name — a support bot or writing tool built on a vendor's model — you're its provider for Article 50, so disclosure and output marking are your duties.
For the full picture, read our Article 50 guide or the text of Article 50. This checker is a first pass, not legal advice.