EU Artificial Intelligence Act transparency rules now in place across Europe

The European Union has started enforcing key transparency requirements under its Artificial Intelligence Act, introducing new obligations for AI developers and businesses

The new rules took effect on August 2 and are designed to make it easier for people to recognise when they are interacting with artificial intelligence and to identify AI-generated or manipulated content.

The new rules under the Artificial Intelligence Act mean that chatbots, AI agents, digital avatars and other interactive AI systems must clearly inform users that they are communicating with artificial intelligence rather than a human.

The rules aim to improve transparency and help prevent confusion or impersonation as AI becomes more widely used in customer service, online platforms and digital communications.

The Artificial Intelligence Act also introduces stricter requirements for AI-generated and manipulated media.

Deepfake images, videos and audio that realistically depict people, places, objects or events must now be clearly identified. Covered content must also include machine-readable markers that enable online platforms and automated systems to detect AI-generated material more easily.

AI-generated text published on matters of public interest must also be disclosed if it has not been reviewed or overseen by a human editor.

The new rules add new responsibilities to providers of general-purpose AI models.

Developers must maintain detailed technical documentation about their models, establish policies to comply with European copyright law and publish summaries describing the data used to train their systems.

The most advanced AI models, particularly the ones considered capable of posing systemic risks, face additional obligations. These include assessing potential threats, implementing safeguards and working closely with regulators to reduce risks related to cybersecurity, manipulation, loss of control and other serious harms.

The Artificial Intelligence Act also bans several categories of AI systems considered to pose unacceptable risks.

These include systems designed to manipulate people in harmful ways, exploit vulnerable individuals or use certain forms of social scoring that could threaten fundamental rights.

Oversight of the new rules will be shared between several regulatory bodies.

The European Commission’s AI Office will supervise providers of general-purpose AI models and certain AI services integrated into the largest online platforms. National authorities will enforce the rules for most other AI systems operating within their countries, while the European Data Protection Supervisor will oversee AI systems used by EU institutions.

The AI Office has also introduced reporting tools that allow individuals, businesses and whistleblowers to report suspected violations confidentially.

To support enforcement, a 60-member Scientific Panel of independent AI experts has begun advising regulators on model testing, safety, evaluation and emerging risks.

While the transparency requirements are now in place, several major parts of the Artificial Intelligence Act have been postponed.

Rules governing high-risk AI systems will now apply from December 2, 2027, while requirements covering AI integrated into regulated products will take effect on August 2, 2028.

The EU also plans to introduce additional prohibitions from December 2, 2026, including bans on AI systems that generate non-consensual sexually explicit material or child sexual abuse content.