OpenAI

OpenAI is best known as the San Francisco lab behind ChatGPT, but a quieter strand of its work runs through the defence world. For most of the company’s life its usage policy explicitly banned “military and warfare” applications. That language was removed in January 2024; the revised policy kept a prohibition on using the service to develop weapons but dropped the blanket bar on military use, opening the door to defence work the company chooses to approve. The change, first reported by The Intercept, marked a clear shift in posture for a firm founded in 2015 under Sam Altman .

The first concrete defence move came in early December 2024, when OpenAI announced a partnership with Anduril focused on counter-unmanned-aircraft systems. The arrangement pairs OpenAI’s models with Anduril’s Lattice software to help detect, assess, and respond to aerial drone threats — framed by both companies around the US–China contest in military AI rather than offensive autonomy.

In June 2025 OpenAI launched “OpenAI for Government,” an initiative consolidating its public-sector deployments — work with the national laboratories, the Air Force Research Laboratory, NASA and others — under a single channel, alongside a government-only ChatGPT. Announced the same day was a contract with the Pentagon’s Chief Digital and Artificial Intelligence Office worth up to $200 million, to prototype frontier AI across both administrative functions and warfighting operations, including agentic workflows. The CDAO later signed comparable agreements with Anthropic , Google and xAI, placing OpenAI among a small group of frontier labs now contracted directly to the Department of Defense. For a company whose public face is a consumer chatbot, it is a striking repositioning — and a reminder that the same general-purpose models sold to the public are increasingly the ones governments want for national-security work.