The AI Act is a European legislative proposal on artificial intelligence that defines the levels of risk associated with the impact of different AI systems on people’s lives and their fundamental rights, from work to health to safety.
The European Union has marked a historic moment in AI regulation with the approval of the AI Act, the first global legislation aimed at regulating the use of AI. This achievement is the result of a long and complex negotiation, known as the “trilogue,” which involved the European Commission, the Council of the European Union, and the European Parliament in a debate lasting over 36 hours.

Historical context: the first global act on AI

The European Commissioner for the Internal Market, Thierry Breton, proudly announced the conclusion of the agreement, emphasising the historical significance of the AI Act. This legislation represents a step forward in AI regulation worldwide, establishing a unique legal framework for the development of reliable and safe artificial intelligence. Ursula von der Leyen, President of the European Commission, underlined how the AI Act aims to guarantee the safety and fundamental rights of individuals and businesses, with the goal of fostering innovation and positioning Europe as a leader in the sector.

The key points of the trilogue

The trilogue represented the culmination of a four-year process during which AI technologies and applications evolved rapidly. One of the main challenges was to balance user safety and the promotion of innovation while respecting fundamental rights and European values. The trilogue addressed several key issues, with a particular focus on:

1. Large AI models (for example ChatGPT)

The agreement establishes a binding yet balanced framework for large AI models, defining transparency requirements for all general-purpose models and imposing stricter rules for those with systemic impacts in the EU Single Market. This two-tier approach is designed to promote innovation along the AI value chain while effectively managing systemic risks.

2. Protection of fundamental rights

The trilogue dedicated significant time to finding a balance between harnessing AI’s potential for law enforcement and protecting citizens’ fundamental rights. The proportionate, risk-based approach allowed for banning the use of AI in violation of EU rights and values. Well-calibrated prohibitions, such as real-time facial recognition, were established with few exemptions, ensuring an effective and proportionate approach to high-risk cases.

3. Fostering innovation

The AI Act serves as a launchpad for EU startups and researchers, promoting innovation through the introduction of regulatory sandboxes and the possibility to test high-risk AI systems in real-world conditions. Future-proof technical standards have been placed at the core of the regulation, ensuring that Europe remains at the forefront of technological developments.

4. Enforcement framework and sanctions

The agreement defined a robust enforcement framework for the AI Act, with market surveillance at the national level and the establishment of a new EU AI Office. Penalties for violations are severe, with fines proportional to the companies’ global annual turnover. The AI Act is set to enter into force two years after approval, with some exceptions for specific provisions.

Conclusions and future prospects

With the approval of the AI Act, the EU positions itself as a pioneer in establishing clear rules for the responsible use of AI. This legislative act is not just a conclusion but the beginning of a new era for responsible and innovative AI development. It offers opportunities to startups and researchers, positioning Europe as a global leader in the race for trustworthy artificial intelligence.