The European Union is set to begin enforcing key provisions of its landmark Artificial Intelligence Act on 2 August, including new rules requiring some AI systems and
AI-generated content to be clearly identified.
Under the rules, chatbots and other interactive AI systems will have to make clear when users are communicating with a machine rather than a person. AI-generated or manipulated images, video and audio – including deepfakes – will also have to be labelled.
Content generated or altered by AI will additionally need machine-readable markings, making it easier for automated systems to identify it.
The European Commission says the measures are designed to reduce deception and manipulation while helping people understand when they are interacting with AI or viewing AI-generated material.
The Commission has also published a first list of more than 180 organisations that have signed its Code of Practice on transparency for AI-generated content.
New enforcement powers
From 2 August, the Commission's AI Office will be able to enforce rules covering providers of general-purpose AI (GPAI) models – systems capable of performing a wide range of tasks and used in products such as AI assistants and agents.
Providers of the most advanced GPAI models, where there is a potential for systemic risks, will face additional requirements. These include measures addressing risks related to chemical, biological, radiological and nuclear incidents, cybercrime, harmful manipulation, threats to fundamental rights and the possibility of AI systems operating beyond human control.
All GPAI providers must document specified information and make it available to authorities and downstream providers. They must also introduce a copyright policy and publish a detailed summary of the material used to train their models.
The new enforcement phase will also cover prohibited AI practices, including systems designed to manipulate people, exploit vulnerabilities in harmful ways or carry out certain forms of social scoring that could undermine people's rights.
Responsibility for enforcing the rules will be divided between the European Commission, national authorities and the European Data Protection Supervisor.
The AI Office will oversee systems supplied by providers of GPAI models, as well as AI systems incorporated into very large online platforms and search engines covered by the EU's Digital Services Act.
National authorities will enforce the rules for other AI systems, while the European Data Protection Supervisor will be responsible for systems used by EU institutions and agencies.
The Commission said effective enforcement would also require member states to appoint and adequately fund their national authorities.
Experts to support enforcement
The AI Office and national regulators will receive scientific advice from a panel of 60 independent AI experts. The panel has recently held its first meeting.
The Commission has also appointed Professor Alessandro Abate of the University of Oxford as its Lead Scientific Adviser. He will advise the AI Office on general-purpose AI, including model testing, evaluation, innovation and adoption.
Abate has previously worked at Stanford University and SRI International and was an assistant professor at TU Delft. His research focuses on the safety, verification and control of AI-enabled systems.
New tools for complaints and whistleblowers
The AI Office has launched several mechanisms for reporting potential violations of the AI Act.
Individuals and organisations can use a complaint tool to report suspected breaches involving AI providers overseen by the AI Office. Employees and others working with AI providers can use a whistleblower channel to report possible violations confidentially.
A separate reporting channel has also been established for downstream companies using general-purpose AI models.
The Commission said information submitted through the reporting tools would be treated confidentially.
Further rules to come
The application of some AI Act requirements has been delayed under the EU's AI Omnibus legislation.
Rules covering high-risk AI systems will apply from 2 December 2027, while rules for high-risk AI systems incorporated into regulated products will begin on 2 August 2028.
New bans on AI systems that generate non-consensual sexually explicit content or child sexual abuse material are due to take effect on 2 December 2026.
The European Commission says the AI Act is intended to create a common framework for developing and using AI safely while giving consumers and businesses greater confidence in the technology. Photo by Jernej Furman from Slovenia, Wikimedia commons.