Short answer: AI Act Article 50 transparency duties have applied since 2 August 2026. People must be told when they interact with an AI system, and AI-generated or manipulated content must be marked or disclosed. On 2 December 2026, two more changes arrive. Generative AI systems that were already on the market before August 2026 must carry machine-readable marking, and two new prohibitions take effect on AI used to create non-consensual intimate imagery and child sexual abuse material. Organisations using chatbots, image generators or AI-written content have only weeks left to check their position.
Key takeaways
- Article 50 covers chatbots, generative AI output, emotion recognition, biometric categorisation and deepfakes.
- 2 December 2026: existing generative AI systems must mark their output in a machine-readable way.
- Two new bans on nudification and child sexual abuse material apply from the same date, with no grandfathering.
- Breaching Article 50 can cost up to €15 million or 3% of worldwide turnover; prohibited practices up to €35 million or 7%.
Using chatbots or generative AI with customers? Consuflix’s AI Act implementation and AI Officer as a Service reviews your AI Act Article 50 disclosures, labelling and vendor set-up before 2 December 2026. Email info@consuflix.com or use our contact page.
What AI Act Article 50 requires
Article 50 of the EU AI Act (Regulation (EU) 2024/1689) sets transparency obligations for certain AI systems, whatever their risk class. The duties are split between providers, who build or place AI on the market, and deployers, who use it.
| Situation | Who | What is required |
|---|---|---|
| AI system interacting directly with people, such as a chatbot or voice agent | Provider | Design it so people are informed they are interacting with AI, unless this is obvious from the context |
| Generative AI creating synthetic audio, images, video or text | Provider | Mark outputs in a machine-readable format so they are detectable as artificially generated or manipulated |
| Emotion recognition or biometric categorisation | Deployer | Inform the people exposed to the system |
| Deepfake image, audio or video content | Deployer | Disclose that the content has been artificially generated or manipulated |
| AI-generated text published to inform the public on matters of public interest | Deployer | Disclose it, unless the content has undergone human review and someone holds editorial responsibility |
The information must be given in a clear and distinguishable manner, at the latest at the time of the first interaction or exposure.
What changes on 2 December 2026
The Digital Omnibus on AI, Regulation (EU) 2026/1744, added a transition date for two topics.
- Marking for existing generative systems. Generative AI systems placed on the market before 2 August 2026 must comply with the machine-readable marking duty in Article 50(2) by 2 December 2026. New systems have had to comply since 2 August 2026.
- Two new prohibited practices. New points in Article 5 prohibit AI systems used to generate non-consensual intimate imagery (“nudification”) and child sexual abuse material. These prohibitions have no grandfathering, so systems already on the market are caught.
Deployers are not off the hook. Even if your vendor handles technical marking, your organisation remains responsible for disclosing deepfakes and informing people about emotion recognition or biometric categorisation that you deploy.
Who should act on AI Act Article 50 now
- Customer service teams using chatbots or voice agents: check that the AI disclosure is shown at the start of the conversation.
- Marketing and content teams using image, video or voice generators: check marking, disclosure of deepfakes and the use of real people’s likeness.
- Publishers and communications teams using AI to draft public-interest text: document human review and editorial responsibility.
- HR, security and retail teams using emotion recognition or biometric categorisation: check whether the use is prohibited (for example, emotion recognition in the workplace is banned under Article 5) before thinking about transparency.
- Product teams building generative features: confirm the marking technique, such as watermarks, metadata or fingerprinting, and test whether it survives normal use.
Transparency under the AI Act sits alongside transparency under the GDPR. If the AI system processes personal data, your privacy notices must also explain it. See our guide to GDPR compliance and our service for Data Protection Officer as a Service.
An AI Act Article 50 checklist before 2 December 2026
- List every chatbot, voice agent and generative AI tool in use, including vendor features.
- For each, record whether you act as provider or deployer.
- Check the AI disclosure wording and where it appears.
- Confirm machine-readable marking with each generative AI vendor, in writing.
- Create a deepfake disclosure rule for marketing and communications.
- Screen all generative tools against the two new prohibitions.
- Train staff who publish content (the Article 4 AI literacy duty already applies). Our article on classifying AI systems helps with steps 1 and 2.
Limits and uncertainty
Technical standards and guidance on marking AI-generated content are still developing, so what counts as sufficient marking may evolve. National authorities will also differ in how they enforce Article 50 at first. This article is general information, not legal advice.
Conclusion
AI Act Article 50 is the part of the AI Act most organisations touch first, because almost everyone now uses chatbots or generative tools. With the 2 December 2026 deadline only weeks away, clear disclosures, confirmed marking and a quick screen against the new prohibitions are practical and achievable steps.
Get your AI transparency right before 2 December 2026. Consuflix’s AI Officer as a Service helps organisations map AI tools, set disclosure rules, check vendors and train staff. Email info@consuflix.com or reach us via our contact page.
Related reading
Sources
- Regulation (EU) 2024/1689 (Artificial Intelligence Act), Articles 5 and 50
- Regulation (EU) 2026/1744 (Digital Omnibus on AI)
- EUAIAct.com: EU AI Act compliance deadlines