Technology
White House keeps voluntary artificial intelligence safety testing framework confidential
The administration finalized review standards in August but keeps them secret, according to CBS News.
The short version
- The White House finalized a voluntary framework in August to evaluate advanced AI models, but it has not shared the criteria publicly.[CBS News]
- Advocacy group Center for Democracy and Technology has formally requested that the administration publish the testing rules.[CBS News]
- OpenAI CEO Sam Altman confirmed submitting the new Astra model for review, describing the evaluation process as productive.[CBS News]
- The ongoing secrecy leaves federal assessment benchmarks and corporate disclosure obligations unclear.[CBS News]
Key facts
- A voluntary federal framework for evaluating advanced artificial intelligence models was finalized by the White House in August.[CBS News]
- The administration's testing protocol remains confidential, leaving specific evaluation benchmarks unrevealed.[CBS News]
- OpenAI submitted its Astra model for evaluation under the framework, according to chief executive Sam Altman.[CBS News]
- The Center for Democracy and Technology has formally pressed the Trump administration to release the evaluation criteria.[CBS News]
What remains uncertain
- The specific evaluation standards and breakthrough reporting criteria required by the administration remain undisclosed.[CBS News]
Sources
Outlet counts describe coverage, not independent confirmation. Reports may share a wire service or original source.
- White House framework for testing new AI models remains hidden as concerns mountCBS News - Top Stories