Glossary · term

International Network for Advanced AI Measurement, Evaluation and Science

The International Network for Advanced AI Measurement, Evaluation and Science is a multilateral forum for government-backed AI institutes and equivalent technical offices. It coordinates research and practices for measuring and evaluating advanced AI. From its 2024 launch until December 2025 it was called the International Network of AI Safety Institutes, which explains the retained AISI-network shorthand.

Regulation2024-05-21Wave 2 · 2024Maturity: 3/5

Origin and context

The initiative was announced at the May 2024 AI Seoul Summit and formally launched at a November 2024 convening in San Francisco. Its ten initial members were Australia, Canada, the European Union, France, Japan, Kenya, the Republic of Korea, Singapore, the United Kingdom, and the United States. In December 2025, members changed the name to emphasize advanced-AI measurement, evaluation, and science, with the UK taking a coordinator role.

Sources: s1, s2, s3

Why it matters

Advanced-model evaluations are difficult to compare when governments use different tasks, reporting conventions, languages, or interpretations. The network offers a venue for technical organizations to exchange methods, conduct joint work, and identify areas of consensus without creating a single global regulator. Its renamed scope matters for current readers: the forum now presents itself around measurement and evaluation science rather than assuming that one test can establish the overall safety of a system.

Sources: s1, s2, s4

Example

Member organizations can compare how automated evaluations are designed and interpreted, document practices on which they agree, and publish open questions that need further research. A shared result can improve comparability across jurisdictions. It does not itself impose a legal requirement on a model provider; any binding obligation must come from the relevant jurisdiction or regulator.

Sources: s1, s4

How it differs

AI safety institutes

AI safety institutes and their successor bodies are national or jurisdictional organizations with their own mandates. The international network connects those bodies for technical cooperation. It has no single national enforcement mandate and should not be described as an institute that independently tests every frontier model.

Maturity and evidence

The forum has a formal mission, a defined multi-jurisdictional membership, convenings, joint work, and published consensus outputs, supporting maturity 3. Its 2025 rename and evolving coordination model show that it is established but still developing. The current canonical name should be used, while the former name remains a necessary historical alias.

Sources: s1, s2, s3, s4

Limits and open questions

The network is a cooperation forum, not a treaty body, certification authority, or supranational regulator. Members differ in powers, resources, and policy priorities, and consensus on evaluation practice does not guarantee identical national decisions. Membership, coordination roles, and terminology can change, so institutional claims require date-stamped official verification.

Sources: s2, s3, s4

References

FACT SHEET: Launch of the International Network of AI Safety InstitutesNational Institute of Standards and Technology · 2024-11-20 · class A
International Network for Advanced AI Measurement, Evaluation and ScienceUK Government · 2025-12-09 · class A
The AI Safety Institute International Network: Next Steps and RecommendationsCenter for Strategic and International Studies · 2024-10-30 · class B
International Network Publishes Consensus Areas on Practices for Automated EvaluationsNational Institute of Standards and Technology · 2026-02-13 · class A

Last updated: 2026-08-27

In the Skills Atlas

This term is also covered in the Skills Atlas as ai risk management skill.