AI giant Anthropic is locked in a dispute with the US Department of Defense (DOD) over how its artificial intelligence models should be deployed. As reported by CNBC, a Pentagon spokesperson told the publication that the company’s $200 million contact with the agency is now ‘under review’. Presently, Anthropic is the only AI company with models running on classified military networks, offering customised tools to national security customers. Anthropic has pushed back against broad military use, seeking assurances that its AI will not be deployed for autonomous weapons or to spy on Americans. As per the report a spokesperson of the company said that Anthropic is having ‘productive conversations, in good faith’ with the Pentagon, emphasising that it remains committed to supporting US national security while ensuring ethical safeguards.
Pentagon’s position
The DOD, however, insists on using Anthropic’s models for ‘all lawful use cases’ without restriction. “If any one company doesn’t want to accommodate that, that’s a problem for us,” said Emil Michael, undersecretary of defense for research and engineering. He warned that limitations could prevent the military from deploying AI in urgent situations.The dispute has drawn political attention. David Sacks, serving as the administration’s AI and crypto czar, accused Anthropic of promoting “woke AI” by resisting certain military applications. The Trump administration has publicly criticized the company in recent months, adding to tensions.
OpenAI and Google agreed to cooperate with Pentagon
Anthropic’s rivals OpenAI and Google also received Pentagon contracts worth up to $200 million last year. Unlike Anthropic, they agreed to allow the DOD to use their models for all lawful purposes, at least within unclassified systems. One company even permitted use across all systems, according to a senior defense official.If Anthropic refuses to comply, the Pentagon could designate it a “supply chain risk”, a label typically reserved for foreign adversaries. Such a move would force vendors and contractors to certify they do not use Anthropic’s models, dealing a major blow to the startup.
