AI Governance Engineer
GitHub

AI Safety Level (ASL)

By Updated

Anthropic's name for tiers of safeguards in its Responsible Scaling Policy. In version 3.4 the policy keeps ASLs only to describe the protections in place today, and sets its requirements through a table of capability thresholds instead 1.

Developed in
the frontier safety frameworks crosswalk
Chapters
ch. 10, Reading List
Source
1 numbered reference, listed below

Where it is used

The term is not used under this name in running prose; the sections listed under "Developed in" treat it.

  • Critical Capability Level (CCL)

    In Google DeepMind's Frontier Safety Framework, a capability level at which a model may pose a heightened risk of severe harm without mitigations; reaching…

  • Responsible Scaling Policy (RSP)

    Anthropic's published policy for catastrophic risks from its models: capability thresholds with planned safeguards, periodic Risk Reports, a Responsible…

Sources

  1. [1] Responsible Scaling Policy, Version 3.4 (effective 2026-07-08; capability-threshold table, Risk Reports, Appendix B on AI Safety Levels). Anthropic. 2026-07-08. https://www-cdn.anthropic.com/files/4zrzovbb/website/0bacdc8440ea96e62a8766d99ebe1d4eea6d5f3a.pdf (verified: primary)

Definitions of legal terms paraphrase the cited text, which governs. Dated statements are as of .

Cite this term

García Aibar, J. (2026). AI Safety Level (ASL). In AI Governance Engineering: The Thesis & Body of Knowledge (v0.5.0), Glossary. https://doi.org/10.5281/zenodo.22956197. https://aigovernanceengineer.com/glossary/ai-safety-level-asl. CC BY 4.0

BibTeX

@misc{aige2026aisafetylevelasl,
  author  = {Jorge García Aibar},
  title   = {{AI Safety Level (ASL)}},
  note    = {Glossary, AI Governance Engineering: The Thesis \& Body of Knowledge, version 0.5.0},
  year    = {2026},
  doi     = {10.5281/zenodo.22956197},
  url     = {https://aigovernanceengineer.com/glossary/ai-safety-level-asl}
}