ChatGPT Coming To an Army Near You

Last week, OpenAI quietly tweaked the wording in its usage policies. That seemingly simple action could have massive consequences.

Most notably, OpenAI removed a previous ban on using its large language models for any military or warfare-related applications. Instead of the ban, the policies now say that OpenAI products can’t be used to “harm yourself or others,” including through weapons development. (This change mostly went unnoticed until The Intercept uncovered it on Friday.)

This narrower restriction could allow the model developer to tap into the lucrative market of defense tech. Under the new policy, there’s plenty the military could still do with OpenAI’s chatbot and LLMs—for example, helping analysts interpret data or write software code. Last August, the Department of Defense launched a task force to explore use cases of generative AI tech like these and others.