When OpenAI was unveiled Its first open-weight model in years this August, tech companies weren’t the only ones paying attention. The release also excited the US military and defense contractors, who saw the opportunity to use them for highly secure operations.
Some military vendors told WIRED that early results show OpenAI’s equipment lags behind competitors in desired capabilities. But they’re still happy that models from a major industry leader are finally an option for them.
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, known as air-gapping. Lilt had previously developed its own AI models or used open source alternatives like Meta’s Llama and Google’s Gemma. But OpenAI’s tools were blocked because they were closed source and could only be accessed online.
New open-weight models from the ChatGPT manufacturer, the gpt-oss-120b and gpt-oss-20b, changed this. Both can run locally, meaning users have the freedom to install them on their devices without requiring a cloud connection. And with access to the models’ weights – key parameters that determine how they respond to different signals – users can tailor them for specific purposes.
OpenAI’s return to the open-source market could ultimately increase competition and create better-performing systems for militaries, health care companies, and others who work with sensitive data. In a recent McKinsey survey of nearly 700 business leaders, more than 50 percent said their organizations use open source AI technologies. Models have different strengths depending on the training process, and organizations often use several together, including open-weighted ones, to ensure reliability in a variety of situations.
Doug Mattie, chief digital and AI officer for the so-called War Department, the name the Trump administration is using for the Defense Department, tells WIRED that the Pentagon plans to integrate generative AI into battlefield systems and back-office functions like auditing. He says some of these applications will require models that are not tied to the cloud. “Our capabilities must be adaptable and flexible,” says Matti.
OpenAI did not respond to requests for comment about how its open source models could be used by the defense industry. Last year, the company reversed a widespread ban on its technology being used for military and warfare applications, a move that prompted criticism from activists worried about the harm AI could cause.
For OpenAI, offering a free and open model can have many benefits. Ease of access can create a large community of experts in its technologies. And because users do not need to sign up as formal customers, they may be able to work with confidentiality, which could prevent OpenAI from facing criticism over potentially controversial customers such as, for example, the military.