Council of the EU adopts landmark AI Act: Setting global standards for AI regulation
On 21 May 2024, the European Council gave its final approval to the pioneering Artificial Intelligence (AI) Act, the first comprehensive AI regulation of its kind worldwide. This significant legislation aims to standardise AI rules across the EU, using a risk-based approach to ensure safety and trust in AI systems. The higher the potential harm to society, the stricter the regulations will be.
Key objectives of the AI Act
- Harmonise AI regulations: Establish uniform rules across the EU for AI systems.
- Promote safe AI: Encourage the development and use of reliable AI systems in both private and public sectors.
- Trust, transparency, and accountability: Highlight the importance of these principles in the development and deployment of AI technologies.
- Protect fundamental rights: Ensure AI systems respect the fundamental rights of EU citizens.
- Stimulate innovation: Boost investment and innovation in AI technologies within Europe.
AI system classification
The Act classifies AI systems by risk:
- Prohibited AI practices: Ban on systems like cognitive behavioural manipulation and social scoring.
- High risk AI: Subject to strict requirements for market access.
- Limited risk AI: Light transparency obligations.
General Purpose AI (GPAI)
The AI act also addresses the use of GPAI models. GPAI models with systemic risks face stringent rules, while those without such risks have minimal requirements.
Governance and enforcement
The new governance structure includes:
- AI office: Ensures consistent rule enforcement across the EU.
- Scientific panel: Provides expert support.
- AI board: Advises on effective application of the Act.
- Advisory forum: Offers technical expertise.
Penalties
Fines for non-compliance are calculated based on a percentage of the company’s global annual turnover or a fixed amount, whichever is higher. SMEs and start-ups are subject to proportional administrative fines.
Transparency and protection of fundamental rights
Before deploying high-risk AI systems, public service providers must assess their impact on fundamental rights. These systems, along with certain users, must be registered in the EU’s database and inform individuals when emotion recognition systems are used.
Innovation support
The Act supports innovation through AI regulatory sandboxes, enabling controlled testing of new AI technologies. The AI act provides for an innovation-friendly legal framework and aims to promote evidence-based regulatory learning.
Implementation timeline
After being signed by the presidents of the European Parliament and the Council, the AI Act will be published in the EU’s Official Journal and come into force 20 days later. Full application of the regulation will begin two years after its entry into force, with some exceptions.
This historic legislation sets a global precedent for AI regulation, striking a balance between technological advancement and the protection of societal interests and ethical standards.
The European Council’s press release can be found here.
Our previous blog post on the AI Act can be accessed here.
