PUBLIC COMMITMENTSRecordUpdated Oct 24, 2026
Public commitments
Security commitments stated in public by model developers and governments
About
This record tracks public security commitments stated by model developers, governments, and international bodies. Each commitment has evidence-backed status updates and source documentation.
Showing 5 commitments: 1 kept, 2 in progress, 0 partially kept, 1 broken, 0 quietly revised, 1 unclear
| Entity | Commitment | Category | Status | Date |
|---|---|---|---|---|
| Anthropic | Pre-deployment safety testing for frontier models | pre-deployment testing | in progress | May 21, 2024 |
| Anthropic | Publish model system cards with capability evaluations | transparency | kept | May 21, 2024 |
| DeepMind | Implement compute governance framework | compute governance | broken | May 21, 2024 |
| UK Government | Establish AI Safety Institute for pre-deployment testing | pre-deployment testing | in progress | Nov 2, 2023 |
| OpenAI | Share safety research and incident information | international cooperation | unclear/unverifiable | Nov 2, 2023 |
Submit a Tip or Correction
Know of a commitment we're missing, or have evidence about a status change? Email [email protected] with details and sources.