EU Enforces Ban on High-Risk AI Systems to Ensure Safety and Ethics

2025-02-03 data

Brussels, Monday, 3 February 2025.
The European Union has enacted legislation banning AI systems with “unacceptable” risks, emphasizing increased AI accountability and affecting tech policies significantly across Europe, especially in the Netherlands.

Critical Risk Categories and Enforcement

As of February 2, 2025, the European Union has established four distinct risk levels for AI systems, with the highest tier now being completely prohibited [1][3]. Systems deemed to have ‘unacceptable risk’ include those used for social scoring, subliminal manipulation, exploitation of vulnerabilities, and biometric surveillance in public spaces [2][3]. Violations of these restrictions can result in substantial penalties of up to €35 million or 7% of annual global turnover, whichever is greater [3].

Implementation Timeline and Industry Response

The EU AI Act, which entered into force on August 1, 2024 [2][3], has garnered significant industry support with over 100 major technology companies, including Microsoft, Google, and OpenAI, signing the EU AI Pact [1]. Notable absences from this voluntary commitment include Apple and Meta, though all companies operating in the EU must comply regardless of their signatory status [1]. The full implementation of the Act will continue to roll out, with the majority of provisions being implemented by August 2, 2026 [2].

AI Literacy Requirements and Compliance

A crucial component of the new regulation is the mandatory AI literacy requirement for organizations [2]. Article 4 of the EU AI Act requires companies to ensure their personnel possess sufficient knowledge and skills to deploy AI systems responsibly [2]. This requirement extends beyond staff to all parties involved in the AI value chain, emphasizing the importance of comprehensive understanding throughout the development and deployment process [2].

Exemptions and Ongoing Developments

While the ban is comprehensive, certain exemptions exist for law enforcement purposes, such as using biometric systems for targeted searches in life-threatening scenarios [3]. The European Commission is expected to release additional guidelines on AI Act prohibitions in early 2025 [3]. Meanwhile, national Market Surveillance Authorities will begin enforcement activities on August 2, 2025 [6], marking another crucial milestone in the implementation of this groundbreaking legislation.

Bronnen

AI regulation EU policy