Credit: Canva The Pentagon’s chief technology officer publicly dismissed a former Anthropic researcher’s viral warning about AI risk and said the department has shifted about 90% of its classified AI workload away from the company, with the rest going by the end of September. The EU AI Act does not cover military or national security uses at all, so the same dispute in Europe would have no regulator to reach. The Pentagon’s chief technology officer has dismissed a former Anthropic researcher’s warning that AI could kill everyone by 2030, Bloomberg reported. Emil Michael called it “fear of data centers, fear of change, all the fears that have coalesced in one well-written tweet“.
Underneath the rhetoric there is a number. Michael told reporters the department has moved about 90% of its classified AI workload off Anthropic, and expects to finish by the end of this month. Palantir’s Maven work went months ago. He also said the Pentagon needs to use software “for all lawful purposes, in an unconstrained way“.
Any harms, he suggested, can be handled by the free market and by talking to industry. That sentence is the entire dispute, and it is not new. TNW published the negotiation emails in July. Dario Amodei held two limits, autonomous weapons and domestic surveillance, and refused the phrase “all lawful uses“ because US law permits the second.
The argument has since moved to court, and Michael has picked his venue. Judge Rita Lin struck down the supply-chain designation in August as retaliation. Michael now calls the pending Washington appeal “more important” and accuses Anthropic of forum-shopping in California. Europe cannot have this argument at all.
The AI Act does not apply to systems put into service or used exclusively for military, defence or national security purposes, regardless of who is carrying out the activity. So a European lab refusing a military use is standing on its contract, not on the regulation Brussels spent five years writing. There is no authority to appeal to and no designation to overturn. Europe’s largest lab is arguing the other side in any case.
Mistral has been building defence models with the German firm Helsing since early 2025. “As long as we have adversaries that are threatening, and they are threatening, we do need to have our own capabilities,” said chief executive Arthur Mensch. He was rebutting Pope Leo XIV, who had called for AI disarmament. Anthropic at least had a court to go to, and the European version of this fight has not been tested anywhere.
