The White House is in advanced talks with the three top AI labs on release standards.
The rules would be voluntary and cover benchmarks, testing timelines, and access.
An announcement is expected soon, per Financial Times reporting.
Washington is trying to set the rules of the road for the most powerful AI models. And it’s writing them alongside the very companies that build them.
What’s Being Negotiated
The Financial Times confirmed the White House is in advanced talks with OpenAI, Google, and Anthropic to finalise voluntary standards for how frontier AI models get released.
An announcement could come as soon as the next week.
The framework would set shared ground rules for the most capable systems. Reporting points to three main areas:
Benchmarks: agreed ways to measure what a model can do
Testing timelines: how long models are evaluated before release
Access rules: who can use advanced models and under what conditions
The word “voluntary” is doing a lot of work here. These would be standards the labs agree to follow, not laws they’re forced to obey. That makes them faster to put in place, but also easier to bend or abandon later.
Why It’s Happening Now
The timing is no accident. This lands in the middle of a heated moment for AI policy.
Consider what’s swirling around it:
Chinese models like Kimi K3 and Qwen3.8 are matching US systems at a fraction of the cost
US labs have accused Chinese firms of stealing model capabilities, which China denies
OpenAI reportedly offered the US government a 5% equity stake, a sign of how tightly it wants to bind itself to Washington
The US government already asked OpenAI to stagger a model release once
Put together, a picture emerges. The US wants to keep its lead in AI, and it sees standards as a way to do that without heavy-handed laws that might slow its own companies down.
There’s a fair criticism worth including. When the companies being regulated help write the regulations, the rules can end up protecting incumbents rather than the public.
Supporters counter that the labs know the technology best, and that voluntary standards now beat rushed laws later.
Either way, the rules that will shape frontier AI are being drafted right now, and only a handful of companies are in the room.
Quick Links: