Source of record
Where this definition comes from
Anthropic Risk Report, August 2026, §2.5, §2.7
“Eight numbered top-level claims; "Claim 1 (models lack strong covert capabilities) is load-bearing for Claims 2–5" (§2.7). Aggregation: "conjunctive, i.e. their strength is roughly given by the product of subclaim strengths; a few are disjunctive ...; some are convergent" (§2.5).”
https://www-cdn.anthropic.com/f61d49fa5596956a5dec75fea0e973bf6a6a8378/Redacted%20Risk%20Report%20August%202026%20.pdf
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 Risk Report, August 2026 | “Eight numbered top-level claims; "Claim 1 (models lack strong covert capabilities) is load-bearing for Claims 2–5" (§2.7). Aggregation: "conjunctive ... a few are disjunctive ...; some are convergent" (§2.5).” | exact confidence: high | Anthropic Risk Report, August 2026 |
| OpenAI OpenAI Preparedness Framework v2 | “Safeguard claims: "Robustness", "Usage Monitoring", "Trust-based Access" (malicious user) and "Lack of Autonomous Capability", "Value Alignment", "Instruction Alignment", "Reliable and Robust System Oversight", "System Architecture" (misaligned model) (App. C).” | close confidence: high | OpenAI Preparedness Framework v2 |
| Google DeepMind Google DeepMind Frontier Safety Framework v3.1 | “"Safety Cases: are an assessable argument showing how severe risks associated with a model's CCLs have been reduced to an acceptable level." (glossary, p.20). Argument-level, not claim-level.” | broad confidence: high | Google DeepMind Frontier Safety Framework v3.1 |
| Meta Meta Advanced AI Scaling Framework v2 | “"safety cases" are named for high-versus-critical determinations; no format given (§4.2.3).” Meta declares 'safety cases' as a trigger for high-vs-critical determinations but does not define the term's format, author, or reviewer -- recorded per the DU rule as a gap, not silently dropped. | none confidence: medium | Meta Advanced AI Scaling Framework v2 |
| UK AI Safety Institute (AISI) UK AISI safety-case vocabulary (discovery) | “UK AISI safety-case vocabulary: claims-arguments-evidence (CAE), GSN, inability argument (discovery) [UV].” Source-document citation tag is {DEU}, which does not appear in the supplied sources table (only {DEV}, a differently-scoped 'Discovery sweep' tag covering SFC/STREAM/AEF-1, is resolvable, and is not specific to UK AISI content). No URL for UK AISI's CAE/GSN/inability-argument material could be resolved without inventing one, so this row is left unverified with the citation url blank pending resolution of {DEU} (confirmed still open in ALIGNMENT-MATRIX.md §6 item 7: 'Not resolved this pass -- not attempted; still open'). Match type 'close' is inferred from context only (CAE explicitly includes 'claims') per the UV translation rule, not confirmed against primary text. | close confidence: low | UK AISI safety-case vocabulary (citation tag {DEU} unresolved in sources table) |
| METR METR | “Frontier Risk Report claims labelled "by [their] importance to the conclusions of our assessment, from low to high importance".” | close confidence: high | METR |
| Frontier Model Forum Frontier Model Forum: Third-Party Assessments | “"Holistic review ... examines the logical structure and evidentiary support for claims about model safety"; "Safety case analysis" inputs are "specific safety claims ..., supporting evidence from evaluations, underlying assumptions, and conditions under which claims hold valid" (2.3).” | close confidence: high | Frontier Model Forum: Third-Party Assessments |
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.
- California SB 53
RL — a truthfulness prohibition on statements about risk, not a definition or instance of a Claim record; recorded as a pointer, not a crosswalk mapping.
California SB 53
