In an undisclosed update to its usage policy, OpenAI has opened the door to military applications for its technology. While the policy previously prohibited the use of its products for "military and war" purposes, that language has now disappeared, and OpenAI has not denied that it is now open to military use.
The Intercept was first to notice the change, which appears to have gone live on January 10th.
In the technology field, as products develop and change, policy wording often undergoes unannounced changes, and OpenAI is obviously no exception. In fact, the company recently announced that its user-customizable GPT would be launched publicly alongside a vague monetization policy, which likely led to some necessary changes.
However, the change in no-militarization policy can hardly be said to be the result of this particular new product. Nor can it be claimed that "military and war" was excluded simply for "clearer" or "easier readability", as OpenAI's statement about this update did. This is a substantive, consequential change in policy, not a restatement of the same policy.
You can read the current usage policy here, and you can read the old policy here. Here are screenshots of the relevant parts:
It's clear that the entire thing has been rewritten, but whether it's more readable is more a matter of personal preference, and having an explicit list of what's not allowed is more readable than replacing it with more general guidelines. But OpenAI policymakers clearly don’t think so, and if this gives them more latitude to interpret favorably or unfavorably a practice that has hitherto been completely unallowable, that’s just a happy side effect.
Although, as OpenAI representative Niko Felix explained, the policy text still prohibits the development and use of weapons across the board, you can see that it was originally listed separately from "Military and Warfare." After all, the military doesn't just make weapons, the weapons are made by others outside of the military.
And it is where these categories do not overlap that OpenAI is supposedly looking for new business opportunities. Not everything the defense establishment does is entirely war-related; as any academic, engineer, or politician knows, the military is deeply involved in a variety of basic research, investments, small business funding, and infrastructure support.
OpenAI’s GPT platform can be of great help to military engineers, who can, for example, summarize decades of documentation on a region’s water infrastructure. How to define and handle the relationship with government and military funds is indeed a difficult problem faced by many companies. Google's "Project Maven" project went too far, but few seemed bothered by the multibillion-dollar JEDI cloud contract. Academic researchers receiving funding from the Air Force Research Laboratory may have no problem using GPT-4, but researchers working on the same project within the Air Force Research Laboratory may not. So where is the line?
Still, OpenAI removed "military and warfare" entirely from prohibited uses, indicating that the company is at least open to serving military customers. I asked the company to confirm or deny this and warned them that the wording of the new policy was clear and any lack of denial would be interpreted as confirmation.