In August 2025, key provisions of the EU AI Act will come into effect. These rules primarily affect providers of general-purpose AI models and introduce new transparency and labeling obligations for companies and public authorities.

Certain provisions of the EU AI Act will apply from 2 August 2025.

Online Training: click here to register with Tecnovy: Registration – AI-Manager Training | 29 Sep – 02 Oct 2025

1. Obligations for Providers of General Purpose AI (GPAI)

• Documentation Requirement: Providers of general-purpose AI models (e.g., large language models like ChatGPT, image AIs like Midjourney) must supply up-to-date technical documentation and make it accessible to downstream AI-system providers.
• Copyright Strategy: Development and disclosure of a policy for complying with copyright law.
• Training Transparency: Creation and publication of a summary of the content used for model training, including the source and type of data.
• Transparency Reports: Publication of the model’s capabilities, limitations, potential risks, and recommendations for integrators.

2. Labeling Requirement for AI-Generated Content

• As of August 2025, companies and public authorities must clearly label all content generated or significantly altered by AI (e.g., texts, images, chatbot responses). The goal is to ensure transparency and prevent deception or misinformation through deepfakes.

3. Governance and Oversight

• Regulatory Authorities: The EU and its Member States must establish relevant authorities and central contact points for companies to consult regarding compliance evaluation.
• EU AI Office: Creation of a European AI Office and a pool of independent scientific experts.
• Sanctions: Starting August 2025, violations can result in substantial fines (up to €35 million or 7% of global annual turnover).

4. Practical Implications for Companies

• Companies using AI models for personnel-related decisions (high-risk AI) must start risk assessment procedures and consult advisory bodies from August 2025.
• As of August 2025, users are not required to report AI use themselves; the focus is on the providers of AI systems.
• In recruitment, with AI-generated content, or when deploying AI chatbots, labeling of AI use is mandatory.

5. Additional Transition Periods

• Not all obligations of the AI Act take effect in August 2025: Further rules, especially those for high-risk AI, will only become mandatory from August 2026 or 2027.