The Pentagon says it barely needs Anthropic anymore.
The Defense Department's chief technology officer publicly rejected a viral AI risk warning issued by a former Anthropic researcher. He also said the Pentagon has already moved about 90% of its classified AI workload away from Anthropic, with the remaining work set to shift by the end of September. The EU AI Act, meanwhile, does not apply to military or national security uses, leaving defense agencies outside that regulatory framework entirely.
This is less about one company's product and more about how governments carve AI risk debates into buckets that don't touch defense work. A researcher can go viral warning the public, and a military procurement office can keep buying and building without missing a beat, because the loudest AI safety fights and the biggest AI contracts don't share a rulebook.
For Anthropic, the practical effect is a shrinking Pentagon contract by month's end, regardless of how the public debate plays out.