White House keeps AI safety rulebook secret from public, smaller labs cry foul

The White House's decision to keep the AI review framework private reflects a broader shift in how frontier AI regulation is being shaped. President Trump ordered officials in June to create the framework, and the administration has spent the last two months developing it—but has chosen not to release it publicly. This opacity is particularly striking because the framework is meant to govern the most powerful AI models in the world and determine which companies qualify as "trusted partners" with early access to them.

The exclusion of open-weight models from the framework creates a structural imbalance. While closed-source leaders like OpenAI and Anthropic submit their models for government review, open-weight competitors—including the Chinese labs Alibaba, DeepSeek, and Moonshot AI—appear to face fewer regulatory obstacles. A source familiar with the process noted that even companies working with open-weight models are unsure whether their products could be submitted or what compliance would mean. This uncertainty may push customers toward using "government-approved" closed-source alternatives, amplifying the market advantage of the firms already in the room with regulators.

Critics across the political spectrum object to the secrecy. Rep. Lori Trahan, co-sponsor of the bipartisan FRONTIER Act, argues governance "belongs in a civilian agency, where it can be seen and questioned, not buried inside the national security apparatus." R Street's Adam Thierer warned the administration is moving toward "a behind-closed-doors de facto licensing regime" that could be "more arbitrary and burdensome" than its predecessor. The central concern is that without transparency, Washington may be quietly deciding who gets a competitive head start in the AI race.