The White House finished a framework for vetting frontier AI models before release, and most of the industry has never seen the actual document.
More than 25 advocacy groups spanning the political spectrum sent a letter asking the administration to publish the voluntary review framework in full. The document, completed August 1, covers how frontier AI models get checked before they reach the public. It was previewed only to a select group of companies ahead of that finalization. Smaller developers and open-source labs were largely left out of the preview.
Pre-deployment review rules set the bar for what counts as a safe launch, and the companies who saw the framework early got a head start on meeting it. That is a real advantage in a market where compliance costs already favor incumbents with legal and safety teams to spare. A framework nobody can read is also a framework nobody outside government can hold anyone accountable to.
Voluntary frameworks tend to stay voluntary right up until they quietly become the industry standard, which is exactly why the letter wants this one out in the open now.