Safety framework version

NIKOLAI proposal (unsourced): the identifier, effective date and change log of the governing safety framework under which an artefact was produced, recorded as a versioned record-type so that an assessment or safeguard can be tied to the exact framework version in force when it was produced.

This is CASRAI's own proposed definition, not a definition any named organisation has agreed to. See what NIKOLAI is and is not.

Source of record

Where this definition comes from

Anthropic Responsible Scaling Policy v3.4, Cover page; Contents; Introduction
“"RSP v3.4", "Effective July 8, 2026" (cover page). Full changelog: "September 19, 2023 (RSP v1.0); October 15, 2024 (RSP v2.0); March 31, 2025 (RSP v2.1); May 14, 2025 (RSP v2.2); February 24, 2026 (RSP v3.0); April 2, 2026 (RSP v3.1); April 29, 2026 (RSP v3.2); May 26, 2026 (RSP v3.3); July 8, 2026 (RSP v3.4)" (Contents page). "We have always intended for our RSP to be a living document" (Introduction).”
https://cdn.sanity.io/files/4zrzovbb/website/0bacdc8440ea96e62a8766d99ebe1d4eea6d5f3a.pdf
Google DeepMind Frontier Safety Framework v3.1, s.5.3, p.17
“"Published: April 17, 2026"; version history "v1.0 (17 May 2024), v2.0 (4 Feb 2025), v3.0 (22 Sep 2025), v3.1 (17 Apr 2026)" (para. of s.5.3, p.17).”
https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/strengthening-our-frontier-safety-framework/frontier-safety-framework_3-1.pdf
Meta Advanced AI Scaling Framework v2, Appendix II, p.44
“Appendix II change log: v2.0 dated April 7, 2026; initial version (Frontier AI Framework) dated February 3, 2025 (para., p.44).”
https://ai.meta.com/static-resource/Meta_Advanced-AI-Scaling-Framework-v2

Crosswalk

How named organisations use this concept

Every row below is a shadow mapping. It is CASRAI's own reading of a published document. No lab, evaluator or regulator named here has declared, endorsed, or been consulted on this mapping. That will change only when an organisation files its own Mapping Declaration — see the non-endorsement policy.

Organisation	Their term, as published	Match	Source
Anthropic Responsible Scaling Policy v3.4 / Risk Report	“"RSP v3.4", "Effective July 8, 2026" (cover page). Full changelog confirmed verbatim across nine versions from RSP v1.0 (19 Sep 2023) through v3.4 (8 Jul 2026) (Contents page). "We have always intended for our RSP to be a living document" (Introduction). The Risk Report records the same threshold and procedure changes introduced in v3.1-v3.4 (§1.3).”	exact confidence: high	Anthropic Responsible Scaling Policy v3.4, effective 8 Jul 2026
OpenAI Preparedness Framework v2 / Frontier Governance Framework	“"Version 2. Last updated: 15th April, 2025" (PF cover). The FGF carries "No date ... anywhere in the PDF" and no version number.”	close confidence: medium	OpenAI Preparedness Framework v2
Google DeepMind Frontier Safety Framework v3.1	“"Published: April 17, 2026"; version history "v1.0 (17 May 2024), v2.0 (4 Feb 2025), v3.0 (22 Sep 2025), v3.1 (17 Apr 2026)" (para. of s.5.3, p.17).”	exact confidence: high	Google DeepMind Frontier Safety Framework v3.1
xAI Frontier AI Framework, 30 Jun 2026	“"Effective Date 30 June 2026"; no version number given; PDF metadata title reads "Privileged/Confidential DRAFT working FRAMEWORK DOC".” This source's PDF metadata /Title reads "Privileged/Confidential DRAFT working FRAMEWORK DOC"; no xAI statement disambiguating draft vs. final status was found (open [VERIFY] item in the source document's register).	narrow confidence: medium	xAI Frontier AI Framework, 30 Jun 2026
Meta Advanced AI Scaling Framework v2	“Appendix II change log: v2.0 dated April 7, 2026; initial version (Frontier AI Framework) dated February 3, 2025 (para., p.44).”	exact confidence: high	Meta Advanced AI Scaling Framework v2
EU Safety and Security chapter	“"Signatories commit to adopting a state-of-the-art Safety and Security Framework ('Framework')" via a three-step process -- creating (Measure 1.1), implementing (1.2), updating (1.3) -- plus notification (1.4) (Commitment 1). Update trigger: "reasonable grounds to believe that the adequacy of their Framework and/or their adherence thereto has been or will be materially undermined, or every 12 months ..., whichever is sooner" (Measure 1.3), with "a changelog, describing how and why the Framework has been updated, along with a version number and the date of change."”	exact confidence: high	EU GPAI Code of Practice, Safety and Security Chapter
California SB 53 SB 53 statute text	“"Frontier AI framework": "documented technical and organizational protocols to manage, assess, and mitigate catastrophic risks" (22757.11(g)); material modification must be published "and a justification for that modification within 30 days" (22757.12(b)(2)).”	close confidence: medium	California SB 53
Safety Framework Cards (discovery) Safety Framework Cards (SSRN 7061798, unread/paywalled)	“"revision protocol" is one of six evaluation dimensions (secondary) [UV].” SFC full text is paywalled on SSRN and could not be read; the {DEV} discovery-sweep tag does not resolve to a single fetchable URL in the provided sources table. Recorded as UV/unverified, not a confirmed match.	none confidence: low	Discovery sweep: Safety Framework Cards (SSRN 7061798, unread/paywalled)

Related, not mapped

Pointers that are not crosswalk claims

These sources mention this concept but do not define or map it clearly enough to count as a crosswalk row — noted here so the research is visible without overstating it as a mapping.

METR
METR maintains a versioned register of published frontier safety policies at /fsp -- a pointer to a tracking resource, not itself a framework-version record.
METR