Source of record
Where this definition comes from
Executive Order 14409, Sec. 3(b)(ii)
“access for a period of up to 30 days before they plan to release such models to other trusted partners”
https://www.whitehouse.gov/presidential-actions/2026/06/promoting-advanced-artificial-intelligence-innovation-and-security/S. 5061 (119th Congress)
“providers give the NSA AI Security Center model access at least 21 calendar days before commerce, including weights”
https://www.congress.gov/bill/119th-congress/senate-bill/5061/text
Crosswalk
How named organisations use this concept
| Organisation | Their term, as published | Match | Source |
|---|---|---|---|
| US Government (Executive Order 14409 / NIST CAISI) Executive Order 14409 / NIST-CAISI | “EO 14409: access "for a period of up to 30 days before they plan to release such models to other trusted partners" (Sec. 3(b)(ii)). CAISI: "pre-deployment evaluations" with no day count.” Also cites {CAISI}. This is a ceiling ('up to') and the clock starts at release 'to other trusted partners', not at public release -- see gap_note below. | exact confidence: high | Executive Order 14409 |
| US Congress (S. 5061, 119th Congress, not enacted) S. 5061 (119th Congress) | “S. 5061 (introduced): providers give the NSA AI Security Center model access "at least 21 calendar days before" commerce, including weights.” A floor ('at least'), the only floor-typed window in the corpus; the bill is introduced, not enacted. | exact confidence: medium | S. 5061 (119th Congress) |
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.
- Anthropic
Global Pacing Level 2: both sides test models before release for "acute risks" in cybersecurity, biology and alignment, possibly via "a global standards body". Risk Report: fewer than about 10 trusted external testers get time-limited unsafeguarded access. No day count.
Amodei, 'We Must Pace the Frontier' - OpenAI
Proposal for a mandatory pre-release CAISI evaluation with a statutory deadline of unstated length, after which the developer may "deploy without penalty" (blueprint p.6). UK AISI received "early access" for "independent, pre-deployment evaluation" of GPT-5.6 (s.9.1.2.6).
OpenAI, 'A Blueprint for a Federal Framework' - Google DeepMind
Hassabis (personal essay, not a Google DeepMind policy document): "Initially, Frontier Labs would voluntarily share models with the Standards Body for review up to 30 days before release." FSF: low-risk external deployments to "a small number of trusted testers".
Hassabis, 'A Framework for Frontier AI' - xAI
Third-party evaluators got an early snapshot (Grok 4.20 model card). CAISI agreement (see GOV row).
Grok 4.20 model card - Meta
"preliminary preparedness assessments" before controlled deployments.
Meta Advanced AI Scaling Framework v2 - EU (GPAI Code of Practice)
No named pre-release third-party window comparable to Hassabis's 'up to 30 days' or EO 14409's. Closest analogue is internal-process timing to the AI Office: Framework must be 'confirmed ... no later than two weeks before placing the model on the market' (Measure 1.1); independent external evaluators implied floor 'at least 20 business days' before the Model Report is filed 'before placing a model on the market' (Commitment 7) -- an implied floor, not a stated one.
EU GPAI Code of Practice, Safety and Security Chapter - Frontier Model Forum
Developers set the timing and depth of engagement.
Frontier Model Forum, Third-Party Assessments
Gap
Both 30-day texts say "up to", which is a ceiling, and EO 14409 ties the clock to release "to other trusted partners", not to public release. S. 5061's "at least" is the only floor. The element needs `windowSemantics` (ceiling or floor) and `clockStartEvent`.
