CoE AI Convention
Council of Europe Framework Convention on AI
Overview
The Council of Europe Framework Convention on Artificial Intelligence, Human Rights, Democracy, and the Rule of Law is the first legally binding international treaty on AI. Adopted in May 2024 and opened for signature in September 2024, it establishes minimum standards for the governance of AI systems to protect human rights and democratic values.
The convention has been signed by major democracies including EU member states, the United Kingdom, the United States, Japan, Canada, and Switzerland. However, as a framework convention, it sets broad principles that each signatory must implement through national legislation or other measures, allowing flexibility in implementation approaches.
The convention is notable for its breadth of support across the democratic world, but its practical impact varies significantly by signatory. The US signature, for example, is largely symbolic without implementing legislation, while EU member states will meet their obligations primarily through the EU AI Act.
Scope
The convention covers AI systems throughout their lifecycle, applying to both public and private sector use. It focuses specifically on AI systems that may impact human rights, democracy, and the rule of law. Signatories may exclude national security and defense applications. The convention applies to the activities of public authorities and, to the extent determined by each party, to private actors.
Key Provisions
Requires parties to ensure that AI systems respect human rights as protected by the European Convention on Human Rights and other applicable international instruments. Includes specific attention to equality, non-discrimination, privacy, and freedom of expression.
Parties must ensure appropriate transparency regarding AI systems, including mechanisms for meaningful human oversight of AI systems that pose significant risks to human rights.
Establishes requirements for accountability in AI development and deployment, including mechanisms for identifying and addressing adverse impacts, and access to effective remedies for persons adversely affected by AI systems.
Requires parties to adopt measures to identify, assess, prevent, and mitigate risks to human rights, democracy, and the rule of law posed by AI systems, using risk-based approaches proportionate to the severity of potential impacts.
Implementation Timeline
May 2024
Convention text adopted by the Council of Europe
September 2024
Opened for signature; initial signatories include EU, UK, US
2025-2026
Ratification process underway in signatory states
TBD
Entry into force (requires minimum number of ratifications)
Ongoing
National implementation measures by each party
Compliance Requirements
- For signatory states: implement national measures to protect human rights in AI
- Establish transparency requirements for AI systems
- Create accountability mechanisms for AI-related harms
- Implement risk assessment processes for AI systems
- Ensure access to remedies for individuals affected by AI systems
- For organizations: monitor national implementation for applicable requirements
Enforcement Mechanism
The convention relies on national implementation for enforcement. Each signatory determines its own implementation approach. A follow-up mechanism through the Conference of the Parties monitors compliance. There are no direct penalties at the international level; enforcement occurs through national legal systems. The monitoring mechanism includes reporting requirements and peer review processes.
Practical Implications
For organizations, the convention's immediate practical impact is limited, as obligations fall on signatory states to implement nationally. However, it signals the direction of international AI governance norms and creates a floor for human rights protection in AI. Organizations should monitor how their home and operating jurisdictions implement the convention's requirements. The convention reinforces the importance of human rights impact assessment as a core AI governance practice.
Relation to EU AI Act
The EU AI Act is the primary mechanism through which EU member states will implement the convention's requirements. The convention's human rights focus aligns closely with the EU AI Act's fundamental rights orientation. The convention is broader in its explicit democratic values focus but less prescriptive in technical requirements. Non-EU signatories like the UK and Japan will need to demonstrate compliance through their own national frameworks, which may or may not match the EU AI Act's level of detail.