This progress builds upon the work that was carried out in previous CAI plenaries and informal meetings. The EU participated actively in the negotiations of the draft convention on the basis of the EU position, in coordination with EU Member States delegates to the CAI. The EU approach for the framework convention aims to provide the parties with sufficient flexibility in the implementation to ensure global reach, while establishing an ambitious, impactful and meaningful legal framework for AI.

The negotiating parties preliminarily agreed on compromise text of many articles, with the notable exception of the scope of the convention, which was left for discussion for the next CAI Plenary on 23-26 January. According to the revised timeline for the negotiations, the convention should be finalised mid-March 2024 and formally adopted on the occasion of the 75^th anniversary of the Council of Europe in May 2024.

Since January 2023, the European Union (EU) is actively participating in the discussions and aiming for a meaningful convention encompassing AI systems that may impact human rights, democracy, and the rule of law, regardless of whether these systems are developed or deployed by public or private actors. The objective is to ensure that this future international instrument is fully compatible with existing relevant EU law on fundamental rights, as well as with the AI Act.

After 37 hours of negotiations, on Friday the 8^th December the EU co-legislators reached a political agreement on the AI Act. The EU AI Act is the world’s first comprehensive AI law. It aims to address risks to health, safety and fundamental rights including democracy, rule of law and environmental protection, while promoting trust and innovation. While most AI systems will pose low to no risk, certain AI systems create risks that need to be addressed to avoid undesirable outcomes. Also, general-purpose AI models that can be integrated in numerous AI systems will be regulated, with more stringent requirements for models that can pose systemic risks. A new AI Office within the Commission will supervise the enforcement of the new rules on general-purpose AI models along with the national market surveillance authorities for all AI systems put into service or made available on the EU market. The legal framework will apply to both public and private actors and will be fully applicable 24 months after entry into force, with earlier application of the rules for general-purpose AI models after 12 months following the entry into force.

To learn more about the EU’s authorisation to negotiate the convention and its position in CAI, read more in our previous blog posts.

To learn more about the AI Act, read more in here.

Article by Lea Koettering - Artificial Intelligence Policy Development and Coordination (CNECT.DDG1.A.2)