EU AI Transparency Rules Take Effect for Chatbots and Deepfakes

European Union transparency rules for artificial intelligence began applying on August 2, 2026, requiring covered providers and deployers to give people clearer notice when they encounter AI systems or synthetic content.

The requirements come from Article 50 of the EU AI Act. They apply to certain interactive and generative AI systems, deepfakes, biometric-categorisation and emotion-recognition tools, and some AI-generated or manipulated text published about matters of public interest.

What providers must disclose

Providers of AI systems that directly interact with people, including chatbots, AI agents and avatars, must make clear that the user is interacting with AI unless that fact is obvious. The notice must be clear and distinguishable from the start of the first interaction.

Providers of covered generative AI systems must also use effective, reliable, robust and interoperable machine-readable marks for AI-generated or manipulated audio, images, video and text. Those marks are intended to help technical systems detect synthetic content.

The rules can reach companies based outside the EU when the output of their AI system is used in the European Union. That gives the requirements potential importance for global technology companies, publishers, advertisers and platforms serving European users.

What users may see

Deployers must clearly disclose deepfakes when people are first exposed to them. A machine-readable mark embedded by the provider is not enough by itself; the disclosure must be understandable to people without special technical tools and will generally need to be visible or audible.

People must also be informed when they are exposed to emotion-recognition or biometric-categorisation systems, whether those systems operate in real time or analyze information afterward.

AI-generated or manipulated text published to inform the public about matters such as politics, public health, public safety, justice, consumer protection, science or the economy must be labelled when it has not received substantive human review or editorial control.

The European Commission says that review must involve deliberate examination by a person with relevant knowledge and professional judgment. Spell-checking, formatting or other superficial checks do not qualify. Text that has been substantively reviewed and controlled by a responsible editorial organization does not require that label under this provision.

Limits and exceptions

The rules do not require every AI-generated image, text or interaction to carry the same disclosure. The guidance includes scope limits for outputs such as source code, short sequences of symbols, machine-to-machine content and some closed-loop industrial uses. Standard editing assistance may also fall outside the marking requirement.

Artistic, creative, satirical and fictional works receive more limited treatment for deepfake disclosures so that notices do not unnecessarily interfere with the presentation of the work.

The EU’s Code of Practice on Transparency of AI-Generated Content is voluntary. It offers a practical way for providers and deployers to demonstrate compliance, but it does not replace Article 50 or create separate mandatory duties.

Who will enforce the rules

Enforcement will mainly rest with national competent market-surveillance authorities rather than being centralized entirely in Brussels. The European AI Office has a limited role for specified systems built on general-purpose AI models, systems where the same entity provides both the model and the system, and certain very large online platforms or search engines. The European Data Protection Supervisor handles systems used by EU institutions, bodies and agencies.

Maximum fines can reach 15 million euros or 3% of a company’s total worldwide annual turnover for the preceding financial year. The European Commission says proportionality may be considered for small and medium-sized companies and small mid-cap companies; the maximum figures are not automatic penalties.

What happens next

A limited transition runs until December 2, 2026, for the marking and detection obligations affecting AI systems placed on the market before August 2, 2026. Content generated before August 2 does not have to be labelled retroactively, although the Commission encourages relevant deployers to do so where possible.

The immediate practical test will be how national authorities interpret the guidance, handle exemptions and address cross-border services. For businesses, the change means mapping where AI systems interact with EU users, generate synthetic media or publish public-interest material—and documenting where meaningful human editorial judgment occurs.

Sources

Discover more from Interactive News

Subscribe to get the latest posts sent to your email.