Sam Altman, chief executive officer of OpenAI Inc., at the AI Impact Summit in New Delhi, India, on Thursday, Feb. 19, 2026.

OpenAI strikes deal with Pentagon

A landmark AI–defense partnership

OpenAI has reached an agreement with the U.S. Department of Defense (DoD) to deploy its artificial intelligence models on the Pentagon’s classified networks. The deal comes just hours after President Donald Trump’s administration blacklisted rival Anthropic from federal contracts, labeling it a “national security supply chain risk”.​

What the OpenAI–Pentagon deal includes

OpenAI CEO Sam Altman announced on X that his company has come to terms with the “Department of War” – his phrasing for the Department of Defense – to bring its models into the Pentagon’s classified environment. Altman emphasized that the DoD has shown “a deep respect” for safety and a genuine willingness to collaborate to get “the best version” of this partnership.​

Under the agreement, OpenAI will build and adapt its AI systems so they can operate securely on sensitive defense networks, with strict safeguards tailored to military use cases. The company is also sending staff to help integrate and support its models inside the Pentagon, ensuring they are deployed, monitored, and updated in a way that aligns with OpenAI’s internal safety standards.​

OpenAI’s red lines on military AI

In an internal memo to employees, Altman stressed that OpenAI shares the same core safety principles on military AI that Anthropic has publicly endorsed. He reiterated two key redlines in his public post: a ban on mass domestic surveillance of citizens and a requirement for human responsibility in decisions about the use of force, including autonomous weapons systems.​

According to Altman, the Pentagon has accepted these conditions and reflected them in law, policy, and in the text of the agreement itself. OpenAI also calls on the Department of Defense to provide transparency on how AI tools are used and audited inside the military, so that the public can understand the boundaries of this cooperation.​

Anthropic blacklisted by Trump administration

The deal with OpenAI landed just after the Trump administration moved to bar Anthropic from all federal contracts, a decision that sent shockwaves through the AI sector. In a formal determination, the administration described Anthropic as a “national security supply chain risk,” effectively cutting the company off from doing business with U.S. government agencies.​

That move immediately reshaped the competitive landscape for defense-related AI work, positioning OpenAI as a preferred partner at a moment when the U.S. government is racing to operationalize advanced models across national security missions. It also raised broader questions in the tech industry about political risk, regulatory exposure, and the level of control Washington can exert over leading AI labs.​

How the Pentagon plans to use OpenAI’s models

While many operational details remain classified, the Pentagon is expected to explore OpenAI’s models for a range of non‑lethal and support functions across the defense ecosystem. These include intelligence analysis, decision support for commanders, logistics planning, cybersecurity operations, and language translation for complex, multilingual data streams.​

The goal is to use large language models and other AI systems to process vast quantities of information faster and more accurately than human analysts could on their own. At the same time, both OpenAI and the DoD insist that humans will stay “in the loop” for any decisions that involve the use of force or significant geopolitical consequences, in line with existing Pentagon doctrine on autonomous systems.​

Safety, ethics, and regulatory pressure

The agreement lands amid intense scrutiny of how frontier AI models might be weaponized or misused in conflict scenarios. Lawmakers, civil society groups, and some AI researchers have warned that poorly governed military AI could increase the risk of accidental escalation, enable new forms of surveillance, or erode democratic oversight over national security decisions.​

OpenAI’s leadership argues that engaging directly with the Pentagon – under strict conditions – is safer than leaving the field to actors who may be less concerned with ethical constraints. At the same time, the company faces internal debate from employees uneasy about any military alignment, as well as external pressure to publish clearer policies and independent audits of its defense work.​

What this means for the future of AI in defense

The OpenAI–Pentagon deal marks one of the clearest signals yet that frontier AI models are moving from labs and commercial products into the core of national security infrastructures. It sets a precedent for how leading AI companies can negotiate safety clauses, red lines, and transparency expectations when working with powerful state actors.​

In the short term, the partnership is likely to accelerate AI adoption inside the U.S. military and set informal standards other contractors will feel compelled to match. In the longer term, it may shape global norms on responsible military AI, as allies and competitors watch how the U.S., OpenAI, and other firms manage the balance between innovation, security, and democratic accountability.​