Safety, Ethics & Governance · beginner · concept 271 of 311 · Updated Nov 4, 2026
Model Specs and Behavior Policies
Public documents that say how a model should behave: what it refuses, whose instructions win in a conflict, and what defaults it keeps. OpenAI's Model Spec, first published in May 2024 and placed in the public domain under CC0, defines a chain of command from Root and System down to Developer, User and Guideline. Anthropic's constitution states the values Claude is trained toward; both labs pair these with usage policies that bind customers.
Top AI contributors on this
- Geoffrey HintonUniversity of Toronto (emeritus)
- Yoshua BengioLawZero / Université de Montréal
- Ilya SutskeverSafe Superintelligence Inc. (SSI)
- Dario AmodeiAnthropic
- Sam AltmanOpenAI
- Jeff DeanDiscovery Loop
Key terms
Learn these first
Where you meet it in the real world
Judging whether a refusal was intended or a bug, writing system prompts that stay inside policy, product and legal review of AI features, public accountability for lab decisions
Videos
OpenAI · YouTube
Dwarkesh Clips · YouTube
Guides and articles
Courses, papers, and more