White House Finalizes Voluntary Frontier-Model Testing Framework on Aug 4 — and Exempts Open-Weight Models Entirely
About a dozen AI companies including OpenAI, Anthropic, Google, and Meta met White House staff on Tuesday August 4, 2026 in a roughly 30-minute session to close two months of negotiation stemming from Trump's June AI cybersecurity executive order; the framework is now final and being implemented. Administered by CAISI inside NIST, it asks developers of covered frontier models — defined as closed-source, state-of-the-art, and national-security-relevant — for a 30-day early access window before public release so the government can run cybersecurity evaluations. Open-source and open-weight models are excluded from the review, which Axios and others frame as a codified competitive asymmetry: five closed labs absorb release friction while open-weight releases face none.
Source
↳ Follow the thread