OpenAI seals classified AI deployment deal with US Dept of War amid Trump

In a major development at the intersection of artificial intelligence and national security, OpenAI has reached a deal with the United States Department of War to deploy its AI models within the department’s classified network, CEO Sam Altman said on Feb 27.

Announced via Altman’s post on X, the agreement comes as tensions between the federal government and rival AI firm Anthropic erupted over differing views on safety protocols and military use of AI.

The deal marks a notable shift in AI engagements with US defense infrastructure, placing OpenAI at the center of sensitive technology deployment.

According to Altman, deployment will proceed under strict safeguards agreed with the Department of War, reflecting OpenAI’s safety priorities.

Two central principles — a ban on domestic mass surveillance and ensuring human responsibility for any use of force, including in autonomous weapon systems — are embedded in the arrangement, he said.

Altman stressed that OpenAI will build technical safeguards to govern model behavior, deploying the systems only on secure cloud networks and with specialised field deployment engineers to manage performance and compliance.

The company has also urged the Department of War to extend similar safety terms to all AI vendors, framing them as industry-wide norms rather than unique concessions.

“We remain committed to serve all of humanity as best we can,” Altman wrote, acknowledging the complexities of global security and technology advancement.

The OpenAI deal was made public against a backdrop of escalating friction between the Trump administration and Anthropic, a rival AI firm.

In recent days, President Donald Trump ordered all federal departments and agencies to immediately halt the use of Anthropic’s technology, after negotiations over safety safeguards collapsed. Trump charged that the company was trying to impose its own terms on the Pentagon’s use of AI.

Following this order, Defence Secretary Pete Hegseth labelled Anthropic a national security supply-chain risk — a designation usually reserved for concerns involving foreign adversaries — effectively barring the company from engaging in commercial activity with the US military.

Anthropic had insisted on retaining strong restrictions against its models being used for domestic mass surveillance or fully autonomous weapon systems, positions that were central to its safety commitments but reportedly at odds with military demands. The breakdown in negotiations cleared the way for OpenAI’s engagement with the classified network.