"They screwed us": Personality clashes sent Anthropic's models offline
TL;DR
Axios reports that Anthropic had to take its top models Mythos and Fable offline after clashing with the Trump administration. The trigger was a warning from Amazon CEO Andy Jassy to Treasury Secretary Scott Bessent that the models could be jailbroken. Administration officials accuse Anthropic of failing to take a cyber executive order seriously enough. On Friday, the government imposed strict export controls, after which Anthropic pulled the models from the internet entirely.
Nauti's Take
This reads less like a clean safety process and more like power politics wrapped in safety language. Anthropic has to answer why a company built around safety messaging appeared so poorly prepared for the political layer.
At the same time, the account leans heavily on anonymous administration sources. Frontier AI oversight should not depend on whether officials feel personally dismissed.
Briefingshow
The story shows how quickly AI safety questions turn into power questions when governments, cloud partners and model makers stop trusting each other. For companies, the warning is clear: technical compliance is no longer enough if political actors believe they were ignored or surprised.
Sources