Source of record
Where this definition comes from
Anthropic Risk Report (August 2026), §4.8
“we do not expect to meet our ambitious industry-wide recommendations ... in time”
https://www-cdn.anthropic.com/f61d49fa5596956a5dec75fea0e973bf6a6a8378/Redacted%20Risk%20Report%20August%202026%20.pdfGoogle DeepMind Frontier Safety Framework v3.1, s.2.2
“We recommend a security level for each CCL, which reflects our assessment of the minimum appropriate level of security the field of frontier AI should apply to models reaching each CCL.”
https://storage.googleapis.com/deepmind-media/DeepMind.com/Blog/strengthening-our-frontier-safety-framework/frontier-safety-framework_3-1.pdf
Crosswalk
How named organisations use this concept
| Organisation | Their term, as published | Match | Source |
|---|---|---|---|
| Anthropic Anthropic Risk Report (August 2026) | “"recommendations for industry-wide safety" tables stated as "capability/usage threshold → substantive standards for model developers" (para., §3.10, §4.8). Automated R&D standard: internal Usage Policy controls binding employees and contractors "up to and including the company's CEO as well as its most privileged technical employees" (Table 3.10.A). "we do not expect to meet our ambitious industry-wide recommendations ... in time" (§4.8).” | exact confidence: high | Anthropic Risk Report (August 2026) |
| Google DeepMind Frontier Safety Framework v3.1 | “"We recommend a security level for each CCL, which reflects our assessment of the minimum appropriate level of security the field of frontier AI should apply to models reaching each CCL." "we believe these recommendations will only be effective if the entire frontier AI field applies them" (s.2.2).” | exact confidence: high | Google DeepMind Frontier Safety Framework v3.1 |
| Demis Hassabis (personal essay) Hassabis, 'A Framework for Frontier AI' | “Frontier Labs "be encouraged to adopt best practices, such as publishing model cards with technical details, maintaining strong internal cybersecurity, vetting key personnel, and providing sufficient resourcing for safety and security research."” Hassabis's substack post is a personal essay, not a Google DeepMind policy document. | close confidence: medium | Hassabis, 'A Framework for Frontier AI' |
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.
- OpenAI
"in order to avoid a race to the bottom on safety, we keep our safeguards at a level more protective than the other AI developer, and share information to validate this claim." (§4.3)
OpenAI Preparedness Framework v2 - California SB 53
Disclosures "consistent with or superior to industry best practice" encouraged (22757.12(c)(4)).
California SB 53 - METR
"This report is descriptive, not prescriptive, and does not represent METR's recommendations."
METR - Frontier Model Forum
"One approach" (a stringent industry-wide baseline) floated, not adopted.
Frontier Model Forum, Risk Taxonomy and Thresholds
Gap
Anthropic and Google both publish recommendations they state they cannot or will not meet unilaterally. NIKOLAI should record `recommendation.selfConformance` (met / not met / expected not to be met) separately from the recommendation itself.
