White House Keeps AI Evaluation Framework Secret
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- White House will not publicly release its voluntary framework for evaluating advanced AI models, with details available only to companies in the process, per three sources cited by Axios
- Meta, Anthropic, Google, OpenAI, Microsoft, and Nvidia sent staffers to the White House Tuesday to review the AI oversight framework, per Reuters and Politico, with Nvidia's inclusion notable as a chipmaker rather than AI lab
- Anthropic models were forced offline by the administration in connection with a rogue AI agent episode that prompted the meeting, per NBC News
- Twelve-plus outlets including Axios, Reuters, Fortune, Wired, and CNN uniformly framed the secrecy as a transparency failure, with Gizmodo declaring 'The White House's New AI Safety Framework Is None of Your Business' and Fortune calling it 'Baffling'
- Lawmakers and researchers including Rep. Lori Trahan, Rep. Greg Casar, Gary Marcus, Stella Biderman, and Nathan Calvin publicly criticized the backroom approach, with Trahan stating 'It's disappointing that they have chosen to block its public release'
Why it matters: Withholding the evaluation criteria from public view means researchers, journalists, and smaller AI developers cannot scrutinize how frontier models will be judged—creating an information asymmetry that favors the handful of companies already in the room as AI capabilities grow more consequential.
