White House framework for testing AI models remains hidden as concerns about threats mount

Washington — The White House has yet to publicly release the voluntary framework for testing frontier AI models that it finalized in August, with no indication of when it might do so, as current and former Anthropic employees issue warnings about AI's potential threat to humans.

Anthropic, the developer of Claude, said in a report released Thursday that it shut down multiple potential efforts by scientists to use Claude to conduct research that could have potentially built biological weapons. 

The White House finalized the framework by the beginning of August, following an executive order President Trump signed in June. The June executive order asked AI companies like Anthropic and OpenAI to voluntarily give the federal government access to their most advanced models up to 30 days before releasing them, with a goal of enhancing the models' security. Mr. Trump pulled the plug on an initial executive order signing because he said he didn't want the federal government to slow down innovation.

The framework that the administration is using to evaluate new models remains confidential, so it's not clear what standards the federal government is asking the companies to meet or whether the firms are disclosing new breakthroughs. Neither the administration nor the companies are required to release the results of the reviews or even say if they participated.

Last week, OpenAI CEO Sam Altman said the company had submitted its powerful new Astra model for review, and called the process "productive." 

"As these models get to a quite significant level of capability, I think the importance of really doing this, closely engaging with the safety institutes in the U.S., the U.K. and elsewhere in the world, will become more important," Altman told Axios.

This week, Anthropic researcher and former OpenAI employee Jacob Coxon resigned from Anthropic, saying on X that "the people building AI earnestly believe that it could kill us all by the end of the decade." Evan Hubinger, another Anthropic employee, agreed, saying, "We really do earnestly believe A.I. could kill all humans!" 

The Center for Democracy and Technology, in a letter to the White House, urged the Trump administration to release its framework for review of frontier models. 

"The public has a right to know what the government's interpretation of the law is on these issues," said Tim Harper, who focuses on election security issues at the Center for Democracy & Technology. "And I think this speaks to a broader lack of transparency we've seen from the administration to disclose internal dialogue and discussions." 

In July, more than 1,000 employees of AI companies posted a statement requesting that the "U.S. government support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development." 

The White House has pressed forward with supporting the development of advanced AI models, despite internal and external warnings from Anthropic, because of the prevailing belief that the U.S. must achieve global dominance in artificial intelligence and outpace China. 

Mr. Trump has expressed strong support for AI development and the building of data centers in the U.S., saying "China could not be happier" about opposition to new infrastructure.

Recent CBS News polling found that a majority of Americans think that AI will take away jobs in the U.S. and nearly two-thirds believe the U.S. government policy will probably not make sure that AI is used in appropriate ways. 

The White House did not respond to repeated requests for comment.