The US government is asking major AI labs to share details of new models 90 days before public release. This voluntary push aims to create a template for future mandatory regulations. The move signals growing pressure on the industry to accept oversight.
The Disclosure Request
Officials want companies to submit model information three months prior to launch, covering capabilities and potential risks. This isn't a binding rule but a framework for how future requirements might work. The 90-day window is the administration's proposed standard timeline for review.
The government argues that waiting until after models hit the market makes safety checks nearly impossible. With powerful systems releasing faster than regulators can react, advance notice could help spot issues before they spread. Current rules don't cover this proactive approach, leaving gaps as AI evolves.
Building the Regulatory Path
By starting with voluntary disclosure, officials hope to test how the process works without forcing compliance. If companies cooperate, it could smooth the way for formal rules later. The administration sees this as the first step toward mandatory standards that might become law in the coming years.
Industry's Next Move
AI firms now face a choice: agree to the disclosure terms or risk heavier regulations if they refuse. Without naming specific companies, the government has reached out to multiple labs building frontier models. How many sign on will likely shape how quickly mandatory rules follow.
The administration hasn't set a deadline for commitments, but expects responses in the coming months as it finalizes next steps.
