White House Keeps AI Safety Testing Framework Confidential

The Trump administration announced that it has finalized a framework for testing new artificial‑intelligence models for safety and cybersecurity risks. The policy, developed over months of consultations with industry leaders, is intended to establish a vetting process for AI systems before they are released to the public. However, the White House has chosen to keep the details of the framework confidential, a decision that critics say undermines transparency and could advantage companies that operate with less public scrutiny.

A private meeting on Tuesday brought together representatives from OpenAI, Anthropic, Meta, Google, Nvidia and Microsoft to review the proposed testing criteria. Sources report that while the volunteer vetting process has been agreed upon, the administration will not publish the policy and will only share the specific testing parameters with a limited group of tech firms. This selective disclosure raises questions about how the framework will be applied across the broader AI ecosystem and whether it will be enforceable for smaller or emerging developers.

The lack of public disclosure has drawn criticism from privacy advocates and lawmakers who argue that a transparent, publicly available framework is essential for accountability and for ensuring that AI systems meet national security and consumer protection standards. The administration has not yet indicated when, if ever, it will release the full policy to the public.