White House Says Voluntary AI Review Framework Is Complete
The White House said its voluntary AI review framework was complete. Reuters’ indexed report names four invited labs; the June 2 order provides for classified cybersecurity benchmarks, without mandatory model licensing.
Correction / update, October 14, 2026: The order’s 30-day access window precedes release to other trusted partners, not necessarily public launch. Classified benchmarks are distinct from the voluntary framework; unsupported comparative and later-event claims were removed.
Axios reports the White House said its voluntary AI framework met its deadline, with industry discussions planned Tuesday.
Reuters’ accessible indexed opening names Meta, Anthropic, OpenAI and Google as invited to Tuesday talks on voluntary government safety testing, citing three sources. The complete Reuters article was not accessible in this review.
What the executive order says#
Executive Order 14409, dated June 2, required a framework within 60 days. Section 3 directs Treasury, War/NSA and DHS/CISA, in consultation with White House officials and Commerce/NIST, to develop a classified cyber-capability benchmarking process and a threshold for covered frontier models.
Its voluntary mechanism allows government access for up to 30 days before release to other trusted partners. The original article incorrectly described this as necessarily before public release. The order expressly says this section does not authorize mandatory government licensing, preclearance or permits for new models.
Limits#
The order classifies the benchmarking process; that does not automatically classify the entire voluntary framework. Completion reported by officials does not establish public availability of its operational details. Unsupported historical-first, EU-comparison and subsequent open-weight-briefing claims have been removed.