OpenAI's Open-Weight Models Are Coming to the US Military

"Lilt, an AI translation company, contracts with the US military to analyze foreign intelligence. Because the company's software handles sensitive information, it must be installed on government servers and work without an internet connection, a practice known as air-gapping. Lilt previously developed its own AI models or used open source options such as Meta's Llama and Google's Gemma. But OpenAI's tools were off the table because they were closed source and could only be accessed online."

"OpenAI's return to the open-source market could ultimately increase competition and lead to better performing systems for militaries, health care companies, and others working with sensitive data. In a recent McKinsey survey of roughly 700 business leaders, more than 50 percent said their organizations use open source AI technologies. Models have different strengths based on how they were trained, and organizations often use several together, including open-weight ones, to ensure reliability across a wide variety of situations."

OpenAI released open-weight models gpt-oss-120b and gpt-oss-20b that can run locally with accessible model weights, allowing installations on private, air-gapped government servers. US military contractors and defense firms view the change as an opportunity to deploy OpenAI technology for sensitive operations despite early evaluations indicating performance lags behind competitors. Organizations often combine multiple models to cover diverse strengths and improve reliability. A McKinsey survey found over half of business leaders use open source AI. Vendors like Lilt previously relied on Meta's Llama, Google's Gemma, or custom models because closed, online-only models were unsuitable for secure, offline environments.

Read at WIRED