Skip to content
DRAFT OPINION (COM(2021)0206 – C90146/2021 – 2021/0106(COD))

On the proposal for a regulation of the European Parliament and of the Council laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) and amending certain Union Legislative Acts

JURI-PA-719827

Committee on Legal Affairs · Rapporteur: Axel Voss

Official HTML ↗ PDF DOCX
✦ In shortFull text
✦ In short · AI summary of this text, generated 26 Sept 2026

Written by a language model from the full text only; every figure comes from the text. The paragraph links open the paragraph a statement rests on — check the text itself before relying on it.

The Committee on Legal Affairs proposes amendments to the Commission's proposed Artificial Intelligence Act. It calls for a risk-based approach, a new definition of AI systems, bans on certain practices, and rules for high-risk AI systems. It introduces concepts of trustworthy AI, general purpose AI systems, and regulatory sandboxes. It also changes obligations for providers, users, importers, and distributors, and adjusts penalties. The amendments aim to support innovation, especially for SMEs and start-ups, and to ensure consistency with existing laws such as the GDPR.

Committee position. The Committee on Legal Affairs proposes amendments to the AI Act to introduce a risk-based approach, support innovation, and strengthen governance. It calls for changes to definitions, obligations, and penalties.

Key points

  1. The Committee on Legal Affairs calls on the Internal Market and Civil Liberties committees to take into account its amendments to the proposed AI Act.
  2. It adds a recital on trustworthy AI and a new article requiring providers to ensure AI systems are lawful, ethical, and robust.
  3. It redefines AI systems based on machine-based inputs and outputs, and adds definitions for general purpose AI, autonomous systems, risk, deep fakes, and regulatory sandboxes.
  4. It bans AI practices that use subliminal techniques or exploit vulnerabilities to cause significant harm, and restricts real-time remote biometric identification by law enforcement.
  5. It classifies high-risk AI systems based on Union harmonisation legislation and critical areas, requiring risk assessments and compliance with essential requirements.
  6. It sets obligations for providers, users, importers, and distributors, including risk management, data governance, technical documentation, logging, and human oversight.
  7. It requires transparency for AI systems interacting with humans and for deep fakes, and mandates disclosure of artificial content.
  8. It establishes AI regulatory sandboxes and measures to support SMEs and start-ups, including priority access and reduced fees.
  9. It re-establishes the High Level Expert Group on AI and strengthens the European Artificial Intelligence Board with a secretariat and binding decision powers.
  10. It sets penalties for non-compliance, with fines up to 20,000,000 EUR or 4% of turnover for prohibited practices, and up to 10,000,000 EUR or 2% for other violations.
  11. It deletes the quality management system requirement and the document retention article, and adjusts conformity assessment procedures.
  12. It amends annexes to update the list of AI techniques, critical areas, and technical documentation requirements.

Who is affected

  • Providers and users of AI systems in the EU, who must comply with new obligations and risk assessments.
  • SMEs and start-ups, which receive support such as priority access to sandboxes and reduced fees.
  • Importers and distributors of AI systems, who must verify conformity and cooperate with authorities.
  • National competent authorities and the European Artificial Intelligence Board, which gain new tasks and powers.
  • Law enforcement authorities, which face restrictions on real-time remote biometric identification.

Figures and deadlines

  • Fines up to 20,000,000 EUR or 4% of total worldwide annual turnover for prohibited AI practices.
  • Fines up to 10,000,000 EUR or 2% of total worldwide annual turnover for other non-compliance.
  • Providers must keep technical documentation for 3 years after placing on the market.
  • Users must keep logs for at least six months.
  • Serious incidents must be reported within 72 hours.
  • The Regulation shall apply from 48 months following entry into force.
  • AI regulatory sandboxes must be established six months prior to the entry into application of this Regulation.