The world's first binding international treaty on artificial intelligence. Adopted by the Council of Europe in May 2024, opened for signature in Vilnius in September 2024. Signed by the EU, the US, the UK, Israel, and other states. AIPIA aligns its Code of Ethics with the Convention's principles.
For most of the past decade, AI governance has consisted of soft-law instruments: OECD AI Principles (2019), UNESCO Recommendation (2021), national AI strategies, voluntary corporate commitments. The Council of Europe Framework Convention is different. It is a binding treaty: states that sign and ratify it accept international legal obligations on the use of AI by public authorities, and — where they so decide — by private actors.
The Council of Europe negotiated the Convention over three years, with participation from its 46 member states and observer states including the US, Canada, Japan, Israel, Australia, Argentina, Costa Rica, Mexico, Peru, and Uruguay. The Convention is open to non-Council-of-Europe states, making it a candidate for the first global framework on AI binding under international law.
The Convention's three foundational pillars — human rights, democracy, and the rule of law — anchor every operational requirement. Parties commit to implementing measures protecting these values throughout the AI lifecycle: design, development, deployment, and decommissioning.
For practitioners, the Convention matters in two practical ways. First, it sets a baseline that is hard to ignore: as more states ratify, the principles become incorporated into national law and procurement rules. Second, it complements the EU AI Act: the Convention provides the human-rights framing, the AI Act provides the operational detail. Together they form the European model of AI regulation increasingly cited as a reference outside the EU.
The Convention is structured around three interlocking commitments. Each is binding on signatory states with respect to public-sector use of AI; parties may extend obligations to the private sector through national law.
AI activities throughout their lifecycle must respect, protect, and fulfil human rights. Discrimination, privacy violations, manipulation of personal autonomy, and other rights infringements caused by AI fall within the Convention's scope and require remedies.
AI use that undermines democratic institutions, electoral integrity, judicial independence, or pluralistic public debate is incompatible with the Convention. Signatories must ensure AI deployment in public services and political processes respects democratic safeguards.
AI activities are subject to law. Authorities must be able to apply existing legal frameworks — administrative, civil, criminal — to harms caused by AI systems. The Convention rejects accountability gaps created by autonomous decision-making.
Parties must establish procedures to assess AI risks and impacts on human rights, democracy, and the rule of law. Assessments are iterative across the AI lifecycle, not a single sign-off at deployment.
AI systems and their use must be identifiable and amenable to oversight. Affected persons should know when they interact with an AI system and be able to obtain meaningful information about its operation.
Persons harmed by AI must have access to effective remedies — judicial, administrative, or other — to challenge decisions and obtain redress. Remedies must be accessible in practice, not only on paper.
Signatories should foster innovation aligned with the Convention's principles, including through controlled testing environments (regulatory sandboxes) and capacity-building for public authorities.
The Convention and the EU AI Act operate at different levels. The Convention sets foundational principles binding signatory states in international law; the AI Act is directly applicable EU legislation with detailed technical obligations, market-surveillance rules, and penalties.
Practically, EU member states that ratify the Convention will satisfy most of their Convention obligations through AI Act implementation. Non-EU signatories — the UK, US, Israel, and others — must translate the Convention into their own legal frameworks, often through new sectoral legislation or amendments to existing administrative law.
AIPIA's Code of Ethics is designed to satisfy both instruments. Members operating across European, UK, and international markets can rely on a single coherent professional standard rather than reconciling parallel frameworks themselves.
The Council of Europe Framework Convention sets the human-rights baseline for AI. AIPIA training, working groups, and Code of Ethics translate it into operational professional practice.