OK, so they might want to review AI models - for what exactly? Is it to make sure that they have been properly lobotomized to become incapable of software security work? Is it to make sure that they only give true and factually accurate information (especially at a time where clearly only the government disseminates facts and truths to the public)? A properly compensated and sufficiently eloquent spin doctor can reframe almost anything as "security".

Given the erratic course of the current US administration, I have absolutely no idea what to actually expect here.

I assumed it was something like holding any new models for ransom. Pay Trump some crypto, and it gets s released. I suppose Trump could work both ends, competitors could bid on delaying someone else's model.