New European Union transparency rules for artificial intelligence systems and generated content have come into force, extending compliance requirements beyond major technology companies to creators, publishers, agencies and other users whose work reaches the EU market.
The rules under Article 50 of the EU AI Act took effect on August 2 and cover a broad range of organisations and individuals. They can also apply to businesses and creators based outside the EU when their AI systems or content are used in the European market.
Non-compliance can result in fines of up to €15 million or 3% of worldwide annual turnover. EU institutions face penalties of up to €750,000.
The transparency requirements cover four main areas. AI systems that directly interact with people must make users aware that they are communicating with an automated system. Providers of systems producing synthetic images, video, audio or text must ensure outputs can be identified in a machine-readable way.
People exposed to emotion recognition or biometric systems must also be informed. Separately, content that has been generated or manipulated by AI and concerns matters of public interest must be clearly identified.
The rules affect technologies including chatbots, voice assistants, AI companions, avatars, robots, coding agents and other automated systems that interact directly with people.
Writers, bloggers, influencers, designers and podcasters also face new responsibilities. AI-generated summaries, rewritten material and substantially altered text may need to carry a disclosure, particularly when covering areas such as public services, health, justice, public security, consumer protection and the economy.
Minor editing, including spelling corrections, grammar checks and basic stylistic improvements, generally does not require disclosure.
Visual content is also covered. Substantial changes to faces, body shapes, colours, objects or events may require labelling. Deepfake images and videos that could be mistaken for authentic representations of real people, places or events must generally be identified.
Audio creators face similar requirements when realistic speech is generated using the voice of a particular person.
The rules also clarify responsibility between companies and individual workers. Where an advertising agency or another organisation controls the use of an AI system, the organisation generally carries the legal responsibility rather than its employees or contractors working under its direction.
Certain exemptions apply to personal, non-professional use, research, scientific activities and some law-enforcement applications. Artistic, creative, fictional and satirical works may receive lighter treatment, although professional use can still trigger disclosure requirements.
The European Commission’s guidance encourages providers to make AI-generated material detectable through machine-readable markings, while users deploying such content must provide a clear disclosure where required.
The new rules mark a significant expansion of the EU’s AI regulatory framework, bringing transparency obligations into everyday content production and digital services across the bloc.
