Overview
- The administration says it met the executive order’s 60‑day deadline and finished the voluntary framework that sets how leading AI developers should engage the government before releasing advanced models.
- The framework’s contents have not been publicly released and the order classifies the benchmarking method and the threshold that determine which models are covered.
- The framework is meant to define rules for government access to models for up to 30 days and to spell out confidentiality, cybersecurity, insider‑risk, intellectual‑property and nondisclosure protections for those reviews.
- A staff‑level meeting organized by the Office of the National Cyber Director will bring major labs, including Google, Anthropic, OpenAI and Meta, to Washington on Tuesday to review the draft and discuss rollout and participation.
- Observers warn that closed drafting and industry input could make the voluntary process a de facto gate for market access, concentrate testing costs with large firms, and leave allies and smaller developers without clarity on how to opt in.