US agencies quietly tell contractors to drop Anthropic AI, raising safety and vendor‑lock questions

On July 28, several posts on US tech forums claimed that companies working with the US government had received directives to discontinue use of Anthropic’s AI products, including the Claude family of large language models. According to these reports, some organizations were told to immediately stop creating new Anthropic accounts or integrations and to begin migrating existing workflows to alternative tools.

No formal public announcement has yet been issued by federal agencies, so the scope of the directive and its legal basis remain unclear. Observers speculate that stricter cybersecurity rules for government contractors, concerns about handling sensitive data on third‑party systems, or procurement policies aimed at avoiding over‑reliance on a single vendor might be driving the move. Forum posters emphasize that the message is not “stop using generative AI altogether,” but rather “do not rely on this specific provider.”

If accurate, the directive could accelerate a shift toward multi‑vendor AI strategies and self‑hosted models in regulated sectors. At the same time, relying on informal guidance and leaked notices risks confusion on the ground. Clearer communication from US authorities about how they assess AI safety, data protection, and vendor lock‑in will be essential for both suppliers and customers in the coming months.