AI Safety Level (ASL)
AI Safety Levels are tiers defined in Anthropic's responsible scaling policy, modelled on biosafety levels, that set the safety and security standards required for models with increasing dangerous capabilities.
What is AI Safety Level (ASL)?
ASL-2 covers models without significant catastrophic-risk capability; ASL-3, first activated in 2025, adds stronger protections against misuse and weight theft.
How it works
Evaluations determine the level a model requires before deployment.
← All Artificial Intelligence & Superintelligence definitions · ← Back to the A–Z definitions