The First AI Law in the World – European Union Comes to a Provisional Agreement on AI Act Draft by Liliana Kotval
The first legislative proposal of its kind in the whole world- the EU Artificial Intelligence Act- was first presented by the EU Commission in April of 2021, and on December 9th, nearly 3 years later, a provisional agreement was finally made on the Act draft. After 3-day talks between the Council presidency, held by Spain until the end of this year, and the European Parliament’s negotiators, the provisional agreement will ensure that AI systems on the European market and used throughout the EU are to be safe and respect fundamental rights and EU values, all the while also aiming to stimulate European AI investment and innovation. Carme Artigas, the Spanish secretary of state for digitalization and AI has stated:
“This is a historical achievement, and a huge milestone towards the future! Today’s agreement effectively addresses a global challenge in a fast-evolving technological environment on a key area for the future of societies and economies.”
The Act will follow the main concept of having a risk-based approach, where the higher the risk, the stricter the rules. Some AI that is deemed unacceptable in terms of risk (such as cognitive behavioral manipulation, emotion recognitions, social scoring, biometric categorization, predictive policing) will be completely banned from the EU. The main new elements of the provisional agreement concluded upon are:
- Better protection of rights through an obligation of deployers of high-risk AI systems to conduct a fundamental rights impact assessment prior to releasing AI systems;
- Extension of the list of prohibitions, however with the possibility to use remote biometric identification by law enforcement authorities in public spaces;
- A revised system of governance with enforcement powers at EU level;
- Rules on high-impact general-purpose AI models that can cause systemic risk in the future.
Furthermore, this provisional agreement clarifies that the regulations are not to be applied to areas outside the scope of EU law and do not affect or compensate national security. The AI Act will not apply to systems used exclusively for military or defense objectives or for the sole purpose of research and innovation. These regulations fall under law enforcement exceptions, where AI is necessary to provide protection. Fundamental rights will still be protected against any misuses of AI systems by law enforcement.
The governance architecture has been agreed upon to have enforcement at EU level at an AI Office within the Commission to oversee the most advanced AI models, contribute to fostering standards and testing practices, and enforce common rules in all member states. A scientific panel of independent experts will advise the AI office. Additionally, an AI Board will comprise of member states’ representatives to remain as a coordination platform and advisory body to the Commission. Finally, an advisory forum for stakeholders, like industry representatives, SMEs, start-ups, civil society, and academia will provide technical expertise to the AI board.
Those that violate the agreed terms of the AI Act will suffer penalties. The fines for such breaks in the law have been calculated based on the percentage of the offending company’s global annual turnover in the previous financial year or a predetermined amount, whichever is higher. Depending on the gravity of the violations, the financial penalties are as follows:
- €35 million or 7% of revenue for violations of banned AI applications
- €15 million or 3% of revenue for violations of the AI Act’s obligations
- €7,5 million or 1,5% of revenue for the supply of incorrect information
A natural or legal person may also make a complaint to the market surveillance authority concerning non-compliance with the Act.
What are the next steps regarding the Act? This provisional agreement should apply 2 years after its entry into force, which, as of now, should be in 2026. Nevertheless, work will continue at the technical level in the next weeks and the Spanish presidency will submit the text to the member states’ representatives (Coreper) for endorsement. Both institutions will need to confirm the entire propositional text before formal adoption by the co-legislators.
- “Artificial Intelligence Act: Council and Parliament Strike a Deal on the First Rules for AI in the World”, European Council, 09 December 2023, https://www.consilium.europa.eu/en/press/press-releases/2023/12/09/artificial-intelligence-act-council-and-parliament-strike-a-deal-on-the-first-worldwide-rules-for-ai/