OpenAI Secures Defense Department Deal Amid Trump's Ban on Anthropic
In a major move within the artificial intelligence industry, OpenAI has reached a deal with the U.S. Department of Defense (DoD) to implement its AI models on classified networks. This agreement comes shortly after President Donald Trump instructed federal agencies to stop using technology from a competing AI firm, Anthropic. Defense Secretary Pete Hegseth further identified Anthropic as a national security risk in the supply chain. These developments highlight the growing tensions between AI companies and the U.S. government over ethical standards and national security concerns.
The conflict began during talks over Anthropic's AI model, Claude, which was under consideration for a contract potentially worth up to $200 million. Anthropic wanted to impose restrictions on its technology, barring its use for mass domestic surveillance or fully autonomous weapons without human oversight. However, the Pentagon sought unrestricted access for any lawful use, seeing these restrictions as a limitation on military operations. After several months of negotiations, including a meeting between Anthropic CEO Dario Amodei and Secretary Hegseth, discussions broke down when Anthropic refused to compromise.
On February 27, 2026, President Trump announced via Truth Social that all federal agencies must "immediately cease" using Anthropic's technology, allowing a six-month period for transition. He stated that the government has no need or desire to continue business with Anthropic, calling the company's terms "a disastrous mistake" that challenged constitutional authority. Secretary Hegseth also announced on X that Anthropic would be classified as a supply-chain risk, a label typically reserved for threats similar to foreign adversaries. This classification prevents military contractors from conducting business with Anthropic, effectively excluding the company from Pentagon-related projects. Hegseth criticized Anthropic's stance as "arrogance and betrayal," asserting that defense decisions are governed by U.S. law, not corporate policies.
In response, Anthropic expressed deep dismay over the decision and pledged to contest the supply-chain risk designation in court, calling it "legally unsound" and a "dangerous precedent." The company reaffirmed its commitment to its principles, emphasizing that no amount of pressure would alter its stance against mass domestic surveillance and fully autonomous weapons. Anthropic also offered to assist in transitioning services during the phase-out, anticipating little effect on non-military clients.
Shortly after these announcements, OpenAI CEO Sam Altman revealed on X that his company had secured an agreement with the DoD—now referred to as the "Department of War" under the Trump administration—to integrate its models into classified systems. Altman emphasized the DoD's "deep respect for safety" and its alignment with OpenAI's principles, which include bans on domestic mass surveillance and requirements for human accountability in the use of force, such as with autonomous weapons. The agreement includes technical safeguards to ensure adherence to these standards and involves deploying OpenAI personnel for oversight on cloud networks. Altman encouraged the DoD to offer similar terms to other AI firms, advocating for resolution through "reasonable agreements" rather than legal disputes.
This series of events represents a critical juncture in the relationship between AI and government, influencing industry standards for ethical AI implementation. Other AI companies, like Google and xAI, maintain similar DoD contracts and have agreed to broader lawful uses. Critics, including former officials, have described the measures against Anthropic as potentially the most severe domestic AI regulation to date, raising concerns about the impact on private-sector freedom in national security matters. These developments also contribute to ongoing discussions about AI's role in modern warfare, further intensified by recent conflicts showcasing automated systems' capabilities.
