The Anthropic-Pentagon saga just took another turn — and this time, the signal is coming from the very top.
What happened
According to Axios, the White House is drafting guidance that would allow federal agencies to bypass Anthropic’s supply-chain risk designation. A draft executive action is reportedly already in circulation. If enacted, agencies — including the Pentagon — could once again access Claude and Mythos without violating internal procurement rules.
Why it matters
Quick recap: Earlier this year, Anthropic refused to loosen Claude’s safety guardrails for military use — no autonomous weapons, no domestic surveillance. The Pentagon’s response was harsh: they classified Anthropic as a supply-chain risk, effectively banning the company. Since then, the Department of Defense has leaned heavily on Google’s Gemini instead.
But not everyone in Washington sees it as black and white. Some voices within the administration consider the standoff counterproductive — especially given Mythos, Anthropic’s most advanced model, which is widely regarded as exceptionally capable in cybersecurity applications. The irony that the most powerful AI model for security tasks is off-limits to security agencies hasn’t gone unnoticed.
What it means for Anthropic
For Anthropic, a softened ban would be a strategic win — without having to compromise on safety. The message: you can have principles and still keep government clients.
At the same time, the situation shows just how political the AI market has become. Anthropic isn’t just a tech company anymore — it’s a piece on a geopolitical chessboard. Google’s $40 billion investment, the Pentagon fight, now the White House stepping in — this level of complexity was unthinkable two years ago.
Whether the executive action actually materializes and what exactly it looks like remains to be seen. But the wind is shifting.
Sources: Axios via Reuters, CNBC — Pentagon expands Google Gemini use