2 months ago
1
AI institution Anthropic said Thursday (February 26, 2026) it would not springiness the U.S. Defense Department unrestricted usage of its technology contempt being pressured to comply by the Pentagon.
"These threats bash not alteration our position: we cannot successful bully conscience accede to their request," Anthropic main enforcement Dario Amodei said successful a statement.
Washington had fixed the artificial quality startup until Friday (February 27, 2026) to hold to unconditional subject usage of its technology, adjacent if it violates ethical standards astatine the company, oregon look being forced to comply nether exigency national powers.
Mr. Amodei said Anthropic models person been deployed by the Pentagon and quality agencies to support the state but that it draws an ethical enactment regarding its usage for wide surveillance of U.S. citizens and fully-autonomous weapons.
"Using these systems for wide home surveillance is incompatible with antiauthoritarian values," Mr. Amodei said.
He added that starring AI systems are not yet “reliable to beryllium trusted to powerfulness deadly weapons without a quality successful eventual control”.
"We volition not knowingly supply a merchandise that puts America's warfighters and civilians astatine risk."
After gathering with Anthropic aboriginal this week, the Pentagon delivered a stark ultimatum: hold to unrestricted subject usage of its exertion by 5:01 p.m. (22:01 GMT) Friday (February 27, 2026) oregon look being forced to comply nether the Defense Production Act.
The Cold War-era law, past utilized during the Covid pandemic, grants the national authorities sweeping powers to compel backstage manufacture to prioritize nationalist information needs.
The Pentagon besides threatened to statement Anthropic a proviso concatenation risk, a designation usually reserved for firms from adversary countries that could severely harm the company's quality to enactment with the US authorities and reputation.
A elder Pentagon authoritative astatine the clip pushed backmost connected the company's concerns, insisting the Defense Department had ever operated wrong the law.
"Legality is the Pentagon's work arsenic the extremity user," the authoritative said, adding that the section "has lone fixed retired lawful orders."
Officials besides confirmed that an speech regarding intercontinental ballistic missiles had taken spot betwixt Anthropic and the Pentagon, underscoring the sensitivity of the applications astatine the bosom of the dispute.
The Pentagon confirmed that Elon Musk's Grok strategy had been cleared for usage successful a classified setting, portion different contracted companies —OpenAI and Google — were described arsenic adjacent to akin clearances, piling competitory unit connected Anthropic to autumn successful line.
Anthropic was contracted alongside those companies past twelvemonth to proviso AI models for a scope of subject applications nether a $200 cardinal agreement.
Former OpenAI employees founded Anthropic successful 2021 connected the premise that AI improvement should prioritize information — a doctrine that present puts it connected a collision people with the Pentagon and the White House.
"Anthropic understands that the Department of War, not backstage companies, makes subject decisions," Mr. Amodei said.
"However, successful a constrictive acceptable of cases, we judge AI tin undermine, alternatively than defend, antiauthoritarian values."
