Aug 1: the US decides which frontier models face a pre-release gate
A June order's voluntary 30-day government preview gets classified criteria this week — so the next flagship's ship date is now partly Washington's call.
Copy markdown
The 30-day gate, in one line
A June 2 executive order lets AI labs give the US government up to 30 days of pre-release access to "covered frontier models" before wider release. It's voluntary and not a license — but in practice it means the next flagship (Opus, GPT, Gemini) can reach you up to a month later than the lab planned.
Friday's deadline, and the bar is classified
Agencies must stand up a classified benchmarking process by August 1 to decide which models count as "covered." The threshold isn't public, so you won't know a model was gated until it's flagged — planning a launch around a specific model's date just got riskier.
You're not only dealing with Washington
The gate lands amid a global compliance wave: China's binding AI-agent rules took effect July 15 with tiered limits on what an agent may decide on its own, and Illinois now mandates third-party safety audits. Ship agents across regions and the deadlines start to overlap.
What to actually do
Treat model availability as a dependency, not a constant. Keep a fallback provider wired up and avoid hard-coding a single frontier model into launch-critical paths — covered releases can now slip, gate, or pause on policy, not just on the lab's roadmap.