Trump-Appointed AI Weapons Supplier Suddenly Changes Course
OpenAI has recently signed an agreement with the US Department of War (DoW) as a supplier of artificial intelligence tools for American military requirements. The agreement was reached after President Trump explicitly halted the use of Anthropic’s AI tools across all federal agencies in the United States.
Previously, Anthropic was the first AI company to operate its model on America’s classified military network. However, contract renewal negotiations failed after the company requested guarantees that its technology would not be used to develop fully autonomous weapons or conduct mass surveillance of US citizens.
Anthropic’s request angered the Trump administration. Trump openly referred to Anthropic as a radical left-wing “woke” company that endangered American lives and national security.
Trump also stated that the government would not allow any company to dictate how the US military fights and wins wars. For its part, Anthropic remained undaunted and planned to sue the government’s decision to place it on the “supply chain risk” list.
Despite tensions between the US government and Anthropic, the Wall Street Journal reported that the US military continued using Anthropic’s AI when attacking Iran on Saturday 28 February local time. The Journal stated that Anthropic’s AI was used for intelligence purposes, target selection assistance, and battle simulations on the battlefield.
OpenAI Changes Course
Regarding the new appointment of OpenAI to replace Anthropic, OpenAI CEO Sam Altman suddenly changed his position. On Monday 2 March local time, Altman stated that OpenAI was coordinating with the DoW to make several changes to their agreement.
“We are discussing with the DoW to make some additions to our agreement. The aim is to ensure our principles are very clear from the outset,” Altman wrote in a post on his personal X account.
Altman stated that one addition to the agreement confirms that the Pentagon has asserted that OpenAI’s services will not be used by the DoW’s intelligence bodies, such as the NSA.
Any service to bodies under the DoW would require further modification to OpenAI’s contract.
For context, the dispute between the US government and Anthropic was originally triggered by the use of the company’s Claude AI in US military operations in January 2026 to capture Venezuelan President Nicolás Maduro.
Anthropic denied involvement and asserted that its terms of service prohibit the use of AI for purposes of violence, autonomous weapons development, or surveillance. Since then, relations between the White House, Pentagon, and Anthropic have continued to deteriorate.
US Defence Secretary Pete Hegseth accused Anthropic of arrogance and betrayal in a post on platform X. He asserted that US soldiers would never be held hostage by the ideological desires of a technology giant.
Hegseth demanded full access without restrictions to all of Anthropic’s AI models for legitimate purposes. However, he also acknowledged that separating the military system from Claude could not be done instantaneously because the technology had been widely deployed.
The Pentagon is said to continue using Anthropic’s services for a maximum of six months to ensure a smooth transition.