The White House is standing up a clearinghouse, called Gold Eagle, that would require AI developers to get government sign-off before bringing partners into a new frontier-model rollout, per CNBC. It grew out of a June executive order and invites developers to give the government up to 30 days of pre-release access to test and evaluate models. Named participants include OpenAI, Anthropic, Microsoft, Google, Cloudflare, and JPMorgan, and the framework touches existing consortia like Anthropic's Project Glasswing and OpenAI's Daybreak.
The distinction to hold is what is binding versus what is voluntary. On paper Gold Eagle is a voluntary information-sharing program. In practice, a government sign-off on who gets early model access functions as a gate whether or not a statute names it one. This extends the precedent set this spring, when Commerce used export-control authority to force a model offline. There the government controlled whether a model could ship. Here it would shape who inside and outside the country gets to use it first. The mechanism is softer and the reach is broader.
The load-bearing detail is legal. Gold Eagle's information-sharing protections lean on liability coverage under the Cybersecurity Information Sharing Act, which expires September 30 unless Congress renews it. That gives the program a clock and a dependency on legislation the administration does not fully control. For the labs, the practical effect is a new gate in the launch process and a standing 30-day government look at unreleased models. For enterprise buyers, model access is becoming something Washington has a say in, which is a new variable in any procurement that depends on getting a model early.
Bottom Line
The government is moving from controlling whether a frontier model ships to controlling who gets to use it. Watch the September 30 CISA deadline, because the program's teeth depend on it.