Countries / International / multilateral
International / multilateral
The Council of Europe AI Convention is signed but not yet in force (EU concluded it in Apr 2026); the UN Scientific Panel and Global Dialogue on AI Governance began work in 2026, alongside voluntary G7/OECD reporting.
In force (6)
-
UN Scientific Panel on AI
40-member independent scientific panel established by consensus resolution to publish an annual, policy-relevant but non-prescriptive assessment of AI opportunities, risks and impacts. Members were appointed by the General Assembly for a three-year term starting 12 Feb 2026; its preliminary report was released 1 Jul 2026.
In forceEffective 12 Feb 2026 -
UN Global Dialogue on AI Governance
Annual intergovernmental and multi-stakeholder dialogue on AI governance, alternating between Geneva and New York, that receives the Scientific Panel's report. The first session was held 6-7 Jul 2026 in Geneva alongside the AI for Good Summit; the 2027 session is set for New York alongside the STI Forum.
In forceEffective 6 Jul 2026 -
G7 Hiroshima reporting framework
Voluntary public reporting framework through which AI developers, deployers and providers disclose how they implement the G7 Hiroshima Code of Conduct (risk management, testing, transparency, incident handling, content provenance). The first round produced 25 organisational reports; version 2.0 broadens participation, with submissions received by 30 Sep 2026 feeding the next review.
In forceEffective 7 Feb 2025 -
International network of AI safety/security institutes
Government network, launched in November 2024, that shares methods for testing and evaluating advanced AI systems; in December 2025 it was renamed to focus on the science of AI measurement and evaluation, with the UK taking the coordinator role. No binding obligations.
In forceEffective 9 Dec 2025 -
International AI Safety Report
Second annual scientific review of the capabilities and risks of general-purpose AI, authored by over 100 experts. It informs governments' frontier-AI policy but creates no obligations.
In forceAdopted 3 Feb 2026 -
AU Continental AI Strategy
Continental strategy endorsed by the AU Executive Council at its 45th Ordinary Session (Accra, 18-19 Jul 2024) calling for an Africa-centric, development-focused approach to AI, unified national strategies, ethical and inclusive governance and regional cooperation. Non-binding guidance for AU member states.
In forceEffective 19 Jul 2024
Adopted, not yet in force (1)
-
Council of Europe AI Convention
First legally binding international AI treaty: parties must ensure AI activities across the lifecycle respect human rights, democracy and the rule of law, with risk and impact management, transparency, oversight and remedies, applying fully to public authorities and, by party choice, to private actors. Opened for signature 5 Sep 2024; signatories include the EU, UK, US, Canada, Japan, Israel, Switzerland, Norway and Ukraine.
Enacted, not yet in forceAdopted 17 May 2024
What the rules require
Obligation types found across International / multilateral's instruments. Filled dot: imposed by a binding instrument. Ring: guidance only.
| Transparency to users | Council of Europe AI Convention, G7 Hiroshima reporting framework | |
| Impact assessment | Council of Europe AI Convention | |
| Human oversight | Council of Europe AI Convention | |
| Incident reporting | G7 Hiroshima reporting framework | |
| Bias and non-discrimination | Council of Europe AI Convention | |
| Frontier model safety | G7 Hiroshima reporting framework, International network of AI safety/security institutes | |
| Security testing | International network of AI safety/security institutes | |
| Government use of AI | Council of Europe AI Convention |
Timeline
- UN Global Dialogue on AI GovernanceTook effect
- UN Scientific Panel on AITook effect
- UN Scientific Panel on AIAdopted
- G7 Hiroshima reporting frameworkAdopted and took effect
- AU Continental AI StrategyAdopted and took effect