The AI Act categorizes AI systems into four risk levels. Unacceptable risk systems, which pose significant threats to safety and rights, are banned, including practices like harmful manipulation and social scoring, effective February 2025. High-risk AI applications, affecting areas such as healthcare and employment, require strict compliance measures before market entry, with rules becoming effective from August 2026 to 2027. Transparency risk mandates clear disclosures for AI interactions, especially for generative AI, with requirements starting in August 2026. Minimal risk systems face no regulations. To support compliance, the Commission introduced guidelines, a voluntary GPAI Code of Practice, and templates for training summaries. Governance and enforcement fall under the European AI Office and EU authorities. The AI Act, effective August 2024, aims to streamline implementation while fostering innovation. Amendments proposed in the Digital Package aim to simplify compliance and enhance support for SMEs, ensuring a balanced approach to AI regulation and development.
