Anthropic has repeatedly pushed for strict guardrails. The company wants written assurances that its AI models will not be used for large-scale domestic surveillance or deployed in fully autonomous weapons systems. Executives argue that without clear protections, the technology could be misused or make critical errors that put Americans at risk.
Defense officials insist that existing U.S. laws already prohibit mass surveillance and unlawful AI use. They argue that additional restrictions imposed by a private company are unnecessary and inappropriate. From their perspective, national defense policy should not be dictated by a tech firm.
Emil Michael, the Under Secretary of Defense for Research and Engineering, said the department has already made written concessions in the latest contract proposal. He emphasized that any use of AI would remain within legal boundaries and that humans would stay involved in key decision-making processes.
“We have humans in the loop in every aspect,” Michael explained, adding that Americans should trust the military to handle the technology responsibly.
The company said the Pentagon’s latest letter failed to address its core concerns. According to Anthropic, little progress was made regarding protections against mass surveillance or guarantees surrounding autonomous weapons. While discussions are ongoing, both sides appear firmly entrenched.
Defense Secretary Pete Hegseth has set a firm deadline: Anthropic must decide by 5:01 p.m. Friday whether it will move forward with the agreement. If the company walks away, the Pentagon has several options. Officials say they could label Anthropic a supply chain risk, forcing its software out of government systems. Alternatively, they could invoke the Defense Production Act to seize access to the technology if deemed critical to national security.
For now, neither side is backing down.
The outcome of this dispute could shape how artificial intelligence is integrated into U.S. military operations — and how much influence private AI companies have over government use of their technology. As the deadline approaches, the showdown highlights a deeper question: who ultimately controls the future of AI in national defense?
