Ultimatum returned to sender. anthropic, one of the leading and most active artificial intelligence startups, will not make available to the Pentagon its systems of AI for military use.
"Threats do not change our position: We cannot in good conscience give in to the Pentagon's demands" for unrestricted access to its technology, said CEO Dario Amodei, thus rejecting the ultimatum from the War Ministry, as Donald Trump has dubbed it, to defuse the impasse that has arisen over the'use of Claudand, its chatbot, for military purposes.
The head-on confrontation between Anthropic and the Pentagon
Claude is the only AI model available inthe army's most classified systems and the one with the best performance in the most sensitive intelligence activities. The ministry led by Pete Hegseth, he asked – but perhaps it would be better to say “he ordered” – a unlimited access to the system. Anthropic, however, refused to completely lift its security measures, proposing to safeguard two areas: the mass surveillance of the Americans and the development of autonomous weapons, that is, capable of striking without human intervention.
The Pentagon asked for more, and a heated argument ensued. In recent days Hegseth imposed an ultimatum to Anthropic: if it had not complied with its requests within two days it would have cancelled the contracts he has with the company, effectively excluding it. Not only that, the minister also threatened to activate the "Defense Production Act”, a rule usually used during national emergencies (the chaos over tariffs demonstrates the “creative” use that the administration makes of the laws ed.) to force Anthropic to provide the requested services. In this second scenario, however, everything could end up in court, with appeals and litigation that could once again reach the Supreme Court. And Trump would risk another defeat.
Anthropic CEO's response
“We believe that, in a limited series of cases, AI can undermine, rather than defend, democratic values”, writes Anthropic CEO Dario Amodei in black and white in a note.
"Claude is widely deployed across the Department of Defense and other security agencies for critical applications such as intelligence analysis, simulation, operational planning, and cyber operations," he said, noting that he has taken decisions to protect US technological primacy, such as blocking the use of Claude by entities linked to the Chinese Communist Party and supporting strict controls on chip exports. Anthropic, he added, has never challenged specific military operations or sought to limit the use of its technology on a case-by-case basis.
The refusal, however, concerns two specific areas: “The use of AI systems for mass domestic surveillance isincompatible with democratic values”, Amodei wrote, while fully autonomous weapons, which exclude humans from the lethal decision-making process, “are not reliable enough today” and could put military and civilians at risk.
According to the company, the Pentagon has asked AI suppliers toaccept “any legal use” by removing such safeguards, threatening to otherwise exclude Anthropic from its systems, designate it as a supply chain risk, or invoke the Defense Production Act to force the restrictions to be removed. “These threats do not change our position“Amodei said. “We remain ready to continue our work in support of U.S. national security,” the CEO concluded.
