The rules for reviewing frontier AI models are secret. The companies that have read them are the big ones.

image via TNW | The heart of techThe Next Web

September 9, 2026, 12:42 PM

• •More than 25 organisations signed a letter asking the White House to publish its voluntary AI security framework for frontier models.
• •The framework governs pre-deployment review of certain powerful AI models and was reportedly completed on 1 August.
• •According to the article, the White House previewed the framework only with a selected group of companies, excluding many smaller and open-source developers.
• •The coalition says secrecy could entrench the largest AI labs by giving them an information advantage on compliance expectations.
• •Protect Democracy has sued the administration under FOIA to obtain the framework and the legal authority behind it.
• •The article notes that officials could argue some evaluation details should stay private for security reasons, even if broader review parameters were disclosed.

More than 25 advocacy organisations from across the political spectrum are asking the White House to publish its voluntary framework for reviewing frontier AI models before deployment. The article says the framework was completed on 1 August and shown only to a selected group of technology companies, while smaller and open-source developers were largely excluded. Critics argue that keeping the rules secret advantages the biggest AI labs because they can prepare for compliance requirements that others can only guess at. The piece also notes a competing security argument for withholding some details, alongside a FOIA lawsuit seeking disclosure of the framework and its legal basis.

Read original article