Following two years of work – a fairly limited time frame for an international treaty – the Convention on Artificial Intelligence (“AI”), the first legally binding international treaty on AI, was approved on May 17, 2024, at the annual meeting of the Council of Europe’s Committee of Ministers.

The Convention consists of 36 articles, grouped into 8 main chapters:

I.  General provisions

II.  General obligations

III.  Principles related to activities within the lifecycle of artificial intelligence systems

IV.  Remedies

V.  Assessment and mitigation of risks and adverse impacts

VI.  Implementation of the Convention

VII. Follow-up mechanism and co-operation

VIII. Final clauses

Summary

The following is a summary of the main issues addressed by the Convention.

Sources

A number of international legal and policy instruments on artificial intelligence were considered for the adoption of the Framework Convention, including the Declaration and Recommendation of the Committee of Ministers of the Council of Europe on Algorithmic Processes and their Impacts on Human Rights, the OECD Principles of AI, as well as the (EU) Regulation on AI (“AI Act”).

Entry into force and adherence

The Convention enters into force after it is opened for signature in Vilnius, Lithuania, on September 5, 2024, at the Conference of Ministers of Justice.

At least five ratifications are required for effective entry into force, including at least three by Council member states.

Adherence, in addition to including the 46 member states of the Council of Europe, is also permitted with regard to non-European countries.

Scope of application

The treaty, which applies to both the private and public sectors, establishes a legal framework aimed at ensuring respect for human rights, the rule of law and democratic legal standards in the use of AI systems throughout the entire life cycle of the systems: design, development, use and decommissioning of AI systems. It facilitates the effective implementation of existing obligations in the face of AI challenges.

Excluded from its scope are activities related to national security, which, however, must comply with international law and democratic values. In summary, the Convention aims to ensure that the use of AI systems is always done responsibly, in the sense of ensuring that artificial intelligence systems are not used to undermine democratic institutions and processes. Victims of human rights violations related to AI systems must have legal remedies. Where there are risks of human rights violations, states can act with moratoriums or bans regarding their implementation.

Risk assessment and monitoring

The Convention – like, moreover, the EU AI Act – takes a risk-based approach by emphasizing consideration of possible negative consequences.

The treaty establishes a monitoring mechanism through a Conference of the Parties (i.e., States Parties) and requires each Party to establish an independent monitoring mechanism to ensure compliance.

It also promotes public awareness, informed debate and multistakeholder consultations on the appropriate use of AI technology.

In summary, the parties must

• ensure accountability for negative impacts,
• ensure that AI systems respect equality (including gender equality, anti-discrimination measures, and data protection rights),
• provide legal remedies for victims of human rights violations related to AI systems,
• establish procedural safeguards by informing individuals about their interaction with AI systems.