AI Safety Level
In Anthropic's Responsible Scaling Policy framework, a graded safety and security standard, such as ASL-2 or ASL-3, that specifies safeguards appropriate to assessed model risks and capabilities, with higher levels generally requiring stronger deployment and model-weight security protections; the detailed requirements have evolved across policy versions.
Metadata
- aliases
- en
- AI Safety Levels
- ASL
- ASLs
- concept_id
- ai_safety_level
- preferred
- en
- AI Safety Level
- relations
- relations
- prerequisite
- status
- active