International AI Treaties and Agreements: The Global Governance Landscape

Council of Europe Framework Convention on AI

The Framework Convention on Artificial Intelligence and Human Rights, Democracy, and the Rule of Law (CETS No. 225) opened for signature on September 5, 2024 in Vilnius. It is the first legally binding international treaty addressing the full AI lifecycle. Key provisions: Article 4 requires parties to adopt measures ensuring AI activities are consistent with human rights obligations; Article 5 mandates risk-based assessment and mitigation; Article 8 requires transparency and oversight; Article 9 addresses accountability and responsibility; and Article 14 establishes a Conference of the Parties for monitoring compliance. The Convention applies to AI systems used by public authorities and, to the extent parties determine, by private actors. Non-Council of Europe states may accede, making it a global instrument. The US, EU, UK, Canada, Japan, Israel, and Australia participated in negotiations.

OECD AI Principles

The OECD Recommendation on Artificial Intelligence (OECD/LEGAL/0449, adopted May 2019, updated 2024) sets five principles: (1) inclusive growth and sustainable development; (2) human-centered values and fairness; (3) transparency and explainability; (4) robustness, security, and safety; (5) accountability. These are complemented by five policy recommendations for governments. Adopted by all 38 OECD members plus 8 non-member adhering countries (46 total), the Principles serve as a common reference point. They are non-binding but have been incorporated by reference into the G20 AI Principles, the EU AI Act recitals, and numerous national AI strategies.

G7 Hiroshima AI Process

The G7 Hiroshima AI Process (2023) produced a Code of Conduct for developers of advanced AI systems, focusing on foundation models and general-purpose AI. The eleven guiding principles include: identifying and mitigating risks throughout the AI lifecycle; identifying and addressing vulnerabilities, incidents, and misuse patterns; reporting capabilities and limitations publicly; investing in robust security controls; developing and deploying reliable content authentication mechanisms; and promoting AI literacy. The Code of Conduct is voluntary but influential: the EU AI Act's GPAI Code of Practice (Article 56) references the Hiroshima principles, and adherence may be used to demonstrate compliance with AI Act obligations.

United Nations Initiatives

The UN General Assembly adopted Resolution 78/265 (March 2024) by consensus, calling for safe, secure, and trustworthy AI systems that promote sustainable development. The UN High-Level Advisory Body on AI published interim recommendations (December 2023) proposing a global AI governance architecture including an international scientific panel, a policy network, and capacity-building for developing countries. UNESCO's Recommendation on the Ethics of AI (adopted November 2021 by all 193 member states) provides ethical principles including proportionality, do no harm, transparency, and human oversight, though it lacks enforcement mechanisms.

Bilateral and Plurilateral Arrangements

Beyond multilateral treaties, bilateral arrangements shape AI governance: the US-EU Trade and Technology Council (TTC) AI Roadmap aligns terminology and promotes compatible risk-based approaches; the US-UK Atlantic Declaration includes AI safety cooperation commitments; the Global Partnership on AI (GPAI, 29 members) conducts technical working groups on responsible AI. The AI Safety Summit series (Bletchley Park 2023, Seoul 2024, Paris 2025) produced declarations on frontier AI safety signed by major AI-developing nations. These instruments lack legal binding force but establish political commitments and shape domestic legislative agendas.

Practical Implications for AI Developers

International treaties and agreements create a convergent normative framework. Organizations developing or deploying AI globally should: (1) monitor ratification of the Council of Europe Convention in their operating jurisdictions, as it will create binding domestic obligations; (2) use the OECD AI Principles as a baseline governance framework accepted across 46 countries; (3) align GPAI system development with the Hiroshima Code of Conduct, as EU AI Act compliance may reference it; (4) track the evolution of UN governance proposals, particularly for developing-country obligations; and (5) participate in national consultation processes on treaty implementation and AI standards development.

This article is for informational purposes only and does not constitute legal advice. Regulatory requirements change frequently — verify current rules with official sources. Built by Sawai Gyoseishoshi Office, Hiroshima, Japan.