Safe AI Forum publishes international best practices for AI Model Specs
natolambert · x · 2026-11-14
Safe AI Forum released "Emerging International Best Practices for AI Model Specs," written by Edward Kembery, Nathan Lambert, Robert McCarthy and a dozen contributors spanning Safe AI Forum, Trillium Labs, CAICT, Shanghai AI Lab, Singapore, Cambridge, and Oxford—covering both Western and Chinese AI ecosystems.
The report notes frontier developers increasingly publish "model specs" describing intended model behavior, yet independent guidance on writing them barely exists. It covers five areas:
- The value of writing a model spec for developers and society
- A taxonomy of model specs for developers to draw on
- Practical guidance on writing and implementing specs
- Eight open problems around specs for safe, desirable behavior
- Recommendations for frontier AI companies, governments, and research/civil society organizations
Related event: Safe AI Forum Releases International Best Practices for AI Model Specs(2 posts)→
More from AGI Musings
- New book repeats the myth that transformers lack recurrence, critic says it ignores Anthropic's interpretability findings — repligate · 2026-11-14
- The human/AI interregnum: nobody can maintain our systems, and AI isn't ready yet — ivan_bezdomny · 2026-11-14
- Rishabh Agarwal on physical AI: RL is 'the dumbest paradigm,' yet it works — ziv_ravid · 2026-11-14
- Recursive self-improvement has no single starting moment — Cosmos Institute essay dissects the intelligence explosion — sebkrier · 2026-11-14
- Gemini helps WSJ reporter crack her 20-year medical mystery — vivnat · 2026-11-14
- 77% of clinical drugs target the same 20% of popular targets—AI could fix this — jacobkimmel · 2026-11-14