The Trump administration has finalized a voluntary AI safety framework establishing a 30-day pre-release review process for advanced closed models from developers like OpenAI and Anthropic. However, officials face intense criticism for keeping the evaluation criteria secret and sharing them only with select firms. To prevent market disparities and address rapid technological growth, the White House plans to expand this framework to open models once they reach frontier capabilities, while state legislators in California step up their own regulatory efforts amid federal inaction.
White House AI safety framework
- ▪The White House AI safety framework establishes a 30-day pre-release review process to test the most powerful artificial intelligence models for national security risks before public deployment.
- ▪The Trump administration finalized a voluntary AI safety framework requiring federal safety testing of advanced closed AI models from developers like Anthropic and OpenAI before public release.
Selective disclosure to AI labs
- ▪The White House has not publicly released its finalized AI safety framework or the specific evaluation criteria and benchmarks used to judge advanced AI models.
- ▪The Trump administration plans to inform only a select handful of large AI companies of the specific evaluation standards and thresholds used in the safety framework.
Transparency concerns
- ▪Critics argue that keeping the AI evaluation rubrics private prevents researchers, startups, and open-weight developers from understanding how their artificial intelligence systems are being judged.
- ▪Former White House Chief Information Officer Theresa Payton criticized the private AI framework, stating that keeping safety tests hidden destroys public trust and creates an insider's club.
Competition concerns
- ▪Trump administration officials privately express concern that a 30-day testing requirement could stifle development and disincentivize domestic companies from developing open-source AI models.
- ▪Critics argue that disclosing the AI safety framework rules only to favored companies effectively picks winners and losers, hamstringing American innovation in the global race against China.
Framework expansion to open models
- ▪Trump administration officials are expanding the AI framework to prevent a two-tier market where enterprises avoid cheaper open models because they lack the federal safety approvals given to closed models.
- ▪White House officials plan to expand the AI safety framework in the coming months to cover open models once they reach the frontier capabilities of Anthropic's Mythos-class or OpenAI's GPT-5.6.
State-level AI regulation
- ▪California State Senator Steve Padilla stated that a lack of coordinated federal action on artificial intelligence has forced state legislators to act with urgency.
- ▪Former Biden White House official Bruce Reed stated that because Washington has failed to pass major tech laws, states like California must lead on fast-moving technology regulation.
Story comments
Loading comments…