Is there any reason they wouldn't rule the same way on DeepSeek training on OpenAI data? After all, one of the big selling points of GPT has been that businesses can freely use the information provided. They're paying for the service, after all. I'd very be interested to know how DeepSeek's usage (very reasonably assuming that they paid for their OpenAI subscription) is any different.
Businesses can’t freely use the information. There are terms of service freely agreed upon by the user which explicitly deny many use cases—training other models is just one. DeepSeek is not an American company nor is their leader in deep with the new administration. It seems far more likely that this will play out like tiktok—they’ll be attacked publicly and banned for national security reasons.
On further reading, I'll grant the first point. Although I wonder if they'll have a technical out—say they distilled from several smaller research companies that had distilled from OpenAI for research purposes, which to my understanding would not constitute a violation of the terms of service.
As for it getting banned, TikTok was banned partly because of credible accounts of it having been used by China to track political enemies. Are we thinking they'll expand the argument on national security to say that any application that transfers data to China is a national security threat? Because that could be a very slippery slope.
And in any case, such a measure seems like it would only bar access to the DeepSeek app. Surely no one could argue that the underlying open source model, if run locally on American soil, could constitute a security threat, right?
