From the journal

EU AI Office Begins AI Act Enforcement as Transparency Rules Take Effect on 2 August 2026

On 2 August 2026, the European Commission's AI Office and national market surveillance authorities commenced enforcement of the EU AI Act. Article 50 transparency obligations now require providers of chatbots, deepfake tools, and AI-generated content systems to disclose AI involvement to users and embed machine-readable marks in AI outputs. Penalties for GPAI model violations can reach up to 3% of global annual turnover.

2 min read

On 2 August 2026, the European Commission's AI Office and national market surveillance authorities began enforcing the EU Artificial Intelligence Act (Regulation (EU) 2024/1689). The AI Office holds primary enforcement authority over general-purpose AI models. National authorities supervise all other AI systems within their jurisdictions. The European Data Protection Supervisor takes on enforcement powers where EU institutions act as providers or deployers.

The transparency obligations in force derive from Article 50 of the AI Act, as amended by the AI Omnibus (Regulation (EU) 2026/1744). Article 50(1) requires providers of AI systems intended to interact with natural persons to ensure those persons are informed they are dealing with an AI, unless the context makes this obvious. Article 50(2) requires providers of AI-generated content systems to embed machine-readable metadata in outputs so downstream detection tools can identify them as AI-generated. The AI Omnibus grants systems already on the market before 7 May 2026 a transitional period to 2 December 2026 to meet the machine-readable marking requirement.

Chatbot operators, text and image generation platforms, synthetic voice services, and deepfake creation tools must now display AI disclosure notices at the point of user interaction. Deployers who build customer-facing products on top of general-purpose AI models share responsibility for ensuring disclosures appear at the user interface layer. The AI Office can request technical documentation, order corrective measures, and impose fines. Fines for GPAI model violations can reach up to 3% of global annual turnover under Article 99.

The obligation under Article 50(1) to disclose AI interaction applies immediately with no grace period. Only the machine-readable marking requirement under Article 50(2) benefits from the transitional deadline of 2 December 2026 for systems already on the market before 7 May 2026. The Commission's Code of Practice on AI-Generated Content Transparency provides non-binding implementation guidance that supervisors are expected to cite in their enforcement assessments.

Licentium advises clients on EU AI Act compliance, including Article 50 disclosure obligations and GPAI model documentation. Contact us to discuss your compliance position. Work we undertake includes AI Act readiness assessments, Article 50 transparency implementation reviews, GPAI model documentation audits, and regulatory sandbox preparation.

Source: European Commission, Press Release IP/26/1714, Commission starts enforcing AI Act rules and new transparency requirements on 2 August, 2 August 2026