Government bodies set up to test frontier models and build the science of evaluation. Several countries established them from 2023 onwards, with agreements to share methods and findings. They are the main public-sector source of independent capability testing.
Technical view
Functions include pre-deployment testing under agreement with developers, publishing evaluation methodology and open-source tooling, and research into dangerous capabilities and safeguard robustness. Their leverage is voluntary access rather than statutory power in most cases, which is the recurring criticism. Some have been renamed or refocused, which is worth checking against current sources rather than assuming continuity.
Related
Part of AI Terminology Index, section: Governance, Law and Regulation.