EU issues guidance on AI transparency rules

The European Commission published guidelines on Monday 20 July to help companies developing or providing AI systems comply with transparency obligations under the AI Act, which will enter into force on 2 August. Providers must tell users when they are interacting directly with AI, including chatbots, AI agents and avatars, and must build reliable machine-readable markings into generated content.

Deployers must disclose AI-generated or AI-modified content, including deepfakes, public-interest material created without human control or editorial oversight, and exposure to AI-based emotion recognition or biometric categorisation. The rules apply to commercial providers and professional users, not to personal or non-professional uses such as AI-generated greeting cards.

The guidance defines interactive AI systems, synthetic content and public-interest AI-generated texts, while setting exceptions for standard editing such as spelling and grammar corrections. Content generated before 2 August will not need retroactive labelling, and systems already on the market before next 2 August get a limited grace period for marking and detection obligations until next 2 December.

The Commission and the European AI Board also approved the code of practice on transparency in AI-generated content, published on 10 July. National market surveillance authorities will mainly oversee compliance, while the European AI Office will intervene in specified cases involving model providers, AI systems, very large online search engines or very large online platforms. Fines may reach up to €15 million or 3% of the previous financial year’s total worldwide annual turnover.