New rules for AI models and systems came into force on 2 August

On 2 August 2026, the European Commission and national authorities began to implement the new provisions of the Artificial Intelligence (AI) Act. These rules stipulate that AI systems designed to interact directly with individuals, such as chatbots, must inform users that they are communicating with an AI system unless it would be obvious to a reasonably well-informed person, considering the circumstances and context of use.

To facilitate their detection, images, videos and audio files that are generated or altered by AI systems, including deepfakes, must now be identified as such in a machine-readable format.

In principle, national market supervisory authorities will be responsible for enforcing the AI Act within the territory of the Member States. However, the AI Office https://digital-strategy.ec.europa.eu/en/policies/ai-office has certain exclusive powers. In particular, it is responsible for supervising AI systems based on general-purpose AI models where the model and system are developed by the same provider. It is also responsible for systems that constitute or are integrated into large platforms and search engines. The European Data Protection Supervisor will oversee the systems used by the institutions of the European Union.

Since February 2025, it has been illegal to develop systems that manipulate people, exploit vulnerabilities or make unfair classifications affecting their fundamental rights. Since then, providers of general-purpose AI models have also been required to document how the models work, adopt policies to ensure compliance with applicable copyright legislation and disclose information about the content used to train these models.

More recently, on 27 July 2026, the Digital Omnibus on AI https://eur-lex.europa.eu/legal-content/PT/TXT/?uri=CELEX:32026R1744 relaxed certain deadlines set out in the AI Act and simplified a number of its rules whilst maintaining robust safeguards for people's safety and fundamental rights. The rules for high-risk systems, as listed in Annex III to the AI Act (including systems used in biometrics, education, employment, and essential public services), will come into force on 2 December 2027.

Meanwhile, the rules relating to AI systems incorporated into products regulated by the legislation listed in Annex I to the same Act (including those relating to the safety of toys, lifts, and medical devices) will take effect on 2 August 2028. The Digital Omnibus on AI also introduced new prohibitions on systems that generate or manipulate non-consensual sexual content or child pornography, which will apply from 2 December 2026.