\n
![](\"https:\/\/gizmodo.com\/app\/uploads\/2026\/03\/Dario-Amodei-Pentagon-Anthropic-V2-1280x853.jpg\")
<\/p>\n
\n<\/p>\n
An executive order reportedly being mulled by President Donald Trump could deepen\u2014or, who knows, resolve?\u2014its ongoing conflict with Anthropic.<\/p>\n
Yesterday I wrote about reports that Trump was working on creating an AI \u201cworking group\u201d by executive order. This group would be made up of government officials and members of the tech industry, and one of its roles might be to devise a review process for unreleased AI models.<\/p>\n
In other words, after promising a light regulatory touch, Trump might be dipping his toe into creating some AI guardrails.<\/p>\n
I noted that the Times\u2019 sources compared Trump\u2019s potential working group to a similar group in the process of being created in the U.K., and that that group was spurred into existence by the revelations of security vulnerabilities brought about by Anthropic\u2019s Claude Mythos Preview model.<\/p>\n
Yesterday afternoon, the AI companies Microsoft, xAI, and Google all signed deals allowing a Biden-created arm of the Commerce Department called the Center for AI Standards and Innovation (CAISI), to inspect their new models prior to release. Anthropic wasn\u2019t included, but it did sign a similar agreement with CAISI under President Biden in 2024.<\/p>\n
But new information reported by Politico about the in-progress order says it may prohibit companies from \u201cinterfering\u201d with government uses of AI. This is according to four of Politico\u2019s seven anonymous sources cited in the story.<\/p>\n
To refresh your memory, Anthropic has been blacklisted by the Pentagon\u2014ostensibly for interfering with the government\u2019s use of its models. It\u2019s a bizarre story with many unanswered questions, but on its face, Anthropic appears to have refused to lift guardrails aimed at preventing the Pentagon from engaging in mass surveillance or full automation of weapons systems, which resulted in a standoff, threats, and ultimately the designation of Anthropic as a supply chain risk, and the requirement that all Pentagon contractors cut all business ties with it.<\/p>\n
It\u2019s not clear what this even is. Language about \u201cinterfering\u201d could either reinforce Anthropic\u2019s pariah status within the Trump Administration, attempt to circumvent it somehow, seek to resolve the matter while saving face, or be purely symbolic. So far the White House has not addressed any of the specific reports about this order, telling Politico discussion before the order is announced is \u201cspeculation.\u201d<\/p>\n<\/p><\/div>\n