BREAKING: Secret US AI evaluation framework has been partly revealed

A secret US AI evaluation framework has been partly revealed, according to Gary Marcus, who wrote that the government responded to a Freedom of Information Act request and lawsuit from Protect Democracy with 132 pages of records. Protect Democracy shared the records with Marcus.

Marcus wrote that the framework concerns evaluating which "frontier" models can be released, and that almost nobody knows what is in it. He also wrote that many people are rightly upset about it.

According to Marcus, the disclosure followed Protect Democracy's FOIA request and lawsuit seeking to find out what the policy is. He said he had told readers about that request the previous week, and that the government responded promptly, which he described as astonishing.

The records were released as 132 pages, and Marcus said he was sharing the first page, which he wrote sets the tone for the rest.

Marcus characterized the situation by writing that even as Trump says on Truth Social that he is violently opposed to regulating AI, he is in fact regulating AI through a secret AI security framework for evaluating which frontier models can be released.

The excerpt did not include the contents of the first page or the remaining records beyond the description of the page as an image.