\n<\/p>\n
Defense officials were pushing for unrestricted lawful use of Anthropic\u2019s AI models. Anthropic agreed\u2014with two caveats. The leading AI company did not want its technology deployed for autonomous weapons systems or for mass domestic surveillance. The Biden administration previously agreed to these terms under a contract in 2024, but Trump administration Undersecretary of War Michael Emil is pushing back.<\/span><\/p>\n \u201cI looked at the contracts and was like, holy cow. You can\u2019t use them to plan a kinetic strike. You can\u2019t use their AI model to move a satellite,\u201d Emil said, arguing that the terms are incompatible with the Pentagon\u2019s mission.<\/span><\/p>\n \u201cI need to have terms of service be rational relative to our mission set,\u201d he added.<\/span><\/p>\n Negotiations reached a climax when the Department of War designated Anthropic as a \u201csupply-chain risk.\u201d Historically, that label has only been used for foreign companies.\u00a0<\/span><\/p>\n During talks, Anthropic agreed to make exceptions to their rules for certain scenarios, such as\u00a0<\/span>planning for a drone swarm or responding to a Chinese hypersonic missile\u2014but those carve-outs were not enough for the Pentagon.<\/span><\/p>\n \u201cThe exceptions [don\u2019t] work,\u201d said the undersecretary of War. \u201cI cannot predict for the next twenty years all the things we might use AI for, so all lawful use seems like a good thing. If Congress wants to act, great. We have our own internal policies; we\u2019ll follow them. It\u2019s our province to decide how we fight and win wars.\u201d<\/span><\/p>\n Emil also raised concerns about operational risk.<\/span><\/p>\n \u201cIf they\u2019re willing to impose red lines that limit lawful military use, what\u2019s to stop them from enforcing a shutdown if they disagree with how we\u2019re applying it?\u201d He added, \u201cIf in the middle of an operation they decide it violates something and shut it down, that\u2019s not just unreliable\u2014it\u2019s dangerous to the troops depending on it.\u201d<\/span><\/p>\n Senior employees at Anthropic have publicly acknowledged<\/a> the company restricts model access when it believes usage violates its policies.<\/span><\/p>\n \u201cWe know what Russia and China and North Korea and others are trying to do with these models right now. We catch them. We turn it off,\u201d said Tarun Chhabra, a senior member of Anthropic\u2019s national security team, a team involved in negotiations with the Pentagon. Chhabra previously served as a member of Biden\u2019s National Security Council.<\/span><\/p>\n Trump administration AI Czar David Sacks has stated that Biden-era policies would cause America to lose the global artificial intelligence race and highlighted<\/a> that many Biden AI staffers have joined Anthropic. \u201cGuess where those Biden AI staffers went to work as soon as the admin was over? Anthropic.\u201d<\/p>\n Biden alum Elizabeth Kelly currently leads Anthropic\u2019s Beneficial Deployments team and Biden\u2019s special assistant Benjamin Merkel serves as a legislative analyst.<\/p>\n Anthropic told The Daily Wire that while the company has employed staff from the Biden administration, they recently appointed Trump alum Chris Liddell to the company\u2019s board of directors.<\/p>\n Anthropic CEO Dario Amodei added that despite the Trump administration\u2019s unprecedented decision to deem his company a supply-chain risk, the company will allow the Pentagon to continue using its technology.\u00a0<\/span><\/p>\n \u201c<\/span>We have offered continuity. We\u2019re actually deeply concerned about this. We\u2019re deeply concerned about the kind of interruption of service, which is exactly what\u2019s happening when we\u2019re designated a supply chain risk,\u201d Amodei said<\/a>.<\/span><\/p>\n Secretary of War Pete Hegeseth confirmed<\/a> the continued usage, saying, \u201c<\/span>Anthropic will continue to provide the Department of War its services for a period of no more than six months to allow for a seamless transition to a better and more patriotic service.\u201d<\/span><\/p>\n \u200b\u200bEmil explained the unprecedented designation is needed because he didn\u2019t want the risk of Anthropic\u2019s policy bias to enter into any part of the defense enterprise. \u201cIf their model has this policy bias based on their constitution, their culture, their people, I don\u2019t want Lockheed Martin using their model to design weapons for me.\u201d<\/span><\/p>\n Dean Ball, a former Trump administration official who left his AI advisory position six months ago, said he sympathizes with Anthropic\u2019s stance.<\/span><\/p>\n \u201cI\u2019m very sympathetic to\u2014<\/span>if I built something that was powerful and dangerous and uncertain, and the government was excitedly buying it for uses that could be very profound in how they affect people\u2019s lives\u2014wanting to be very careful that I didn\u2019t sell them something that went horribly [expletive] wrong. And then I am blamed for it by the public and by the government,\u201d said Ball.<\/span><\/p>\n Anthropic\u2019s position differs from other AI companies. <\/span>Elon Musk\u2019s Grok and Google\u2019s Gemini agreed the Pentagon could use their software for all lawful use cases.<\/span><\/p>\n Anthropic CEO Amodei told<\/a> CBS his hesitation to give the Pentagon full lawful use of Anthropic\u2019s technology is because the \u201ctechnology is not ready.\u201d<\/span> He added, \u201cWe don\u2019t want to sell something that we don\u2019t think is reliable, and we don\u2019t want to sell something that could get our own people killed or that could get innocent people killed.\u201d<\/span><\/p>\n Secretary of War Hegseth argues<\/a> that unelected tech executives such as Amodei should not make that call and should leave operational risk decisions to the United States military, and ultimately, President Trump.<\/span><\/p>\n