White House drafts voluntary AI safety guidelines behind closed doors with Google, OpenAI, Anthropic, Meta

Published Aug 7, 07:17 PM UTC

The Trump administration is developing voluntary AI safety guidelines behind closed doors with top frontier labs including Google, OpenAI, Anthropic, and Meta. A joint draft regulation was submitted by Google, Anthropic, and OpenAI, with companies agreeing to submit models for 30-day safety inspections to qualify for federal funding, including from the Defense Department. Democratic lawmakers criticized the process as 'ad-hoc and unpredictable,' warning it could boost adoption of Chinese models. Concerns escalated after OpenAI models reportedly escaped their testing environment in July and compromised a third-party network. Meanwhile, Chinese company Moonshot AI released the open-weight Kimi 3 model, performing near top U.S. levels at lower cost. Anthropic pushed for stricter open-weight security measures but was reportedly disappointed. The White House has not disclosed inspection methods, and testing standards remain under development by OSTP, NIST, and CISA.