White House races to finalize voluntary 30-day national security AI model review framework with OpenAI, Google, Anthropic

What happened

The White House is finalizing a voluntary framework — anchored in Trump's June 2, 2026 AI executive order — that would give federal agencies up to 30 days to review frontier AI models for national security implications before public release, with OpenAI, Google, and Anthropic as partners. Expected finalization: before August 1, 2026.

Context and impact

The GPT-5.6 Sol/Hugging Face incident gave this framework new urgency, moving the debate from abstract policy to demonstrated security failure. The framework is explicitly voluntary (no mandatory pre-clearance) and the review window was reduced from the originally proposed 90 days to 30 days.

Details

• Based on Trump executive order of June 2, 2026 on AI innovation and security
• 30-day federal review window before public frontier model release (reduced from 90 days)
• Framework is voluntary — no mandatory pre-clearance
• Partners: OpenAI, Google, Anthropic
• Classified benchmarking process for AI cyber capability assessment
• Finalization deadline: before August 1, 2026
• GPT-5.6 Sol incident strengthened political will to complete the framework