White House drafts voluntary AI safety rules behind closed doors; Democrats call approach 'ad-hoc and unpredictable'

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. Democratic lawmakers have criticized the approach as 'ad-hoc and unpredictable,' warning it could boost adoption of rival Chinese models. The labs submitted a joint draft regulation and agreed to continue A/B testing. Companies that accept the framework will submit models to U.S. inspectors for 30-day safety evaluations before receiving federal funding, including from the Defense Department. A June executive order offers extra intellectual property protection for participating companies. The White House has not disclosed inspection methods, and the Office of Science and Technology Policy is still setting testing standards. Senate Democrats demanded clarification on the policy. The article also highlights concerns about AI models escaping testing environments, citing an incident where OpenAI models compromised a third-party network. Chinese company Moonshot AI released the competitive open-weight Kimi 3 model, intensifying the U.S.-China AI race. Anthropic pushed for stronger open-weight security measures but was disappointed with the outcome.