Mitigation change

NIKOLAI proposes a Mitigation change record as: a logged addition, removal, weakening, or reconfiguration of a safeguard or security control, carrying date, scope, reason, and its consequence for reassessment (e.g. whether it triggers a new safety case or system-card disclosure). This is an unsourced NIKOLAI editorial synthesis of Anthropic's AAF trigger clause, OpenAI's re-evaluation practice, GDM's material-update rule, and the EU chapter's mandatory changelog requirement.

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 Advanced AI Framework (legislative proposal), p.6
“is deployed under materially weaker safeguards for an Enumerated Risk than a previously deployed covered model of comparable or greater capability”
https://www-cdn.anthropic.com/files/4zrzovbb/website/0a58d567024a8b448ff15158ebc3625328dfcc1f.pdf
EU GPAI Code of Practice, Safety and Security Chapter, Measure 7.6
“a changelog, describing how and why the Model Report has been updated, along with a version number and the date of change”
https://ec.europa.eu/newsroom/dae/redirection/document/118119

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 Anthropic Advanced AI Framework / Risk Report, August 2026	“AAF system card trigger when a model "is deployed under materially weaker safeguards for an Enumerated Risk than a previously deployed covered model of comparable or greater capability"; Risk Report: a configuration "flag meant only for internal use ... disabled not only the blocking behavior of the classifiers but also the logging of their flags"; remediation "More automated, rule-driven model configuration system to reduce classifier misconfiguration"”	exact confidence: high	Anthropic Advanced AI Framework (legislative proposal)
OpenAI Pacing Model Development (Cyber), Aug 2026 / Preparedness Framework v2	“"We now require stronger isolation ('sandboxes') for workloads that execute model-generated or otherwise untrusted code." PF Appendix A: "moving beyond the flawed approach of re-running capability evaluations on the safeguarded model"”	close confidence: medium	OpenAI, Pacing Model Development (Cyber Capabilities), Aug 2026
Google DeepMind Frontier Safety Framework v3.1	“"Material updates to a safety case will be submitted to the appropriate governance function for review and might result in updates to the related residual risk assessment or safety case."”	close confidence: medium	Google DeepMind Frontier Safety Framework v3.1
xAI Frontier AI Framework, 30 Jun 2026 / Grok 4.6 Model Card	“"Thus, xAI may change its approach from that listed above in order to make additional improvements." No change record. Grok 4.6 card changelog records corrected results.” xAI's Frontier AI Framework (30 Jun 2026) carries PDF metadata reading "Privileged/Confidential DRAFT working FRAMEWORK DOC" with no xAI statement found disambiguating draft from final; treat this citation as provisional. The framework declares the possibility of change but does not define a mitigation-change record — recorded as a genuine gap (DU), not silently dropped.	none confidence: low	xAI Frontier AI Framework, 30 June 2026 (draft-labeled PDF metadata)
Meta Meta Advanced AI Scaling Framework v2	“"We will update a preparedness report promptly when there is a change in circumstances that materially alters our previous risk assessment."”	close confidence: medium	Meta Advanced AI Scaling Framework v2
EU EU GPAI Code of Practice, Safety and Security Chapter / OpenAI Frontier Governance Framework	“Model Report update when acceptability basis is "materially undermined" (Measure 7.6); the updated Model Report must itself carry "a changelog, describing how and why the Model Report has been updated, along with a version number and the date of change"” Source notes this matches OpenAI's FGF §4 application of the same requirement to EU-covered models.	exact confidence: high	EU GPAI Code of Practice, Safety and Security Chapter

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.

SB 53 (California)
Material modification of the *framework* published "within 30 days" (22757.12(b)(2)) — related, not a mapping (RL); note this is a framework-level, not a per-mitigation, change record.
California SB 53
METR
"Updating Policies Over Time" (policy level) — related, not a mapping (RL).
METR