Source of record
Where this definition comes from
Anthropic Responsible Scaling Policy v3.4, §3.1, Scope and Timing
“"A Risk Report will cover the risks of Anthropic's models and actions as of a given coverage date, which will be no more than 30 days prior to the publication of the report." It covers "all publicly deployed models as of the coverage date, as well as any internally deployed models as of the coverage date" meeting an escalation test; "We will publish a Risk Report every 3-6 months."”
https://cdn.sanity.io/files/4zrzovbb/website/0bacdc8440ea96e62a8766d99ebe1d4eea6d5f3a.pdf
Crosswalk
How named organisations use this concept
| Organisation | Their term, as published | Match | Source |
|---|---|---|---|
| Anthropic Responsible Scaling Policy v3.4 / Risk Report | “"A Risk Report will cover the risks of Anthropic's models and actions as of a given coverage date, which will be no more than 30 days prior to the publication of the report." It covers "all publicly deployed models as of the coverage date, as well as any internally deployed models as of the coverage date" meeting an escalation test (high-stakes misalignment or automated R&D risk "significantly exceed[ing] those posed by models covered by a prior Risk Report"), abbreviated "in-scope models"; "We will publish a Risk Report every 3-6 months." This report: "Coverage date: July 15, 2026".” Also under the "coverage" false-friend label (see source §4): Anthropic and GDM elsewhere use "coverage" to mean the scope of what a safeguard catches, and FMF uses it to mean evaluation coverage -- distinct senses from "coverage date". | exact confidence: high | Anthropic Responsible Scaling Policy v3.4, effective 8 Jul 2026 |
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
Cadence, not an as-of date: "We will in any event determine whether to update the Model Report for our most capable frontier models once every six months" (FGF §4, p.16); the GPT-5.6 card flags in-line additions ("Added August 3, 2026").
OpenAI Frontier Governance Framework - xAI
Card revision date: "Revision: 2026-08-17" with a dated changelog (cover and p.1) -- a document-revision date, not a report coverage date.
Grok 4.6 model card - Meta
Timing only: "We will publish a preparedness report for each closed or open Frontier AI release in a timely manner in connection with the deployment" (§2.2.1, p.7).
Meta Advanced AI Scaling Framework v2 - California SB 53
Transparency report "(C) The release date of the frontier model" (22757.12(c)(1)); incident report "(1) The date of the critical safety incident" (22757.13(a)) -- event dates, not a report coverage date.
California SB 53 - METR
Engagement window: investigation scoped over "June 26 - July 13, 2026" (para.) -- an evaluator's own engagement window, not a lab's coverage date.
METR, OpenAI/Hugging Face incident investigation blog
Divergence
Where sources materially disagree
"Coverage" is a false-friend label (see source §4): Anthropic and GDM also use it for safeguard scope, and FMF uses it for evaluation coverage -- neither sense should be merged with this date field.
Gap
Anthropic is the only source that separates coverage date from publication date. Every incident disclosed between the two dates (Anthropic records UK AISI's Mythos 5 report as occurring "after the coverage date of this report") needs this field to be placed correctly.
