Last week, OpenAI debuted the new version of its video generator Sora. While the videos look better than they did before, there are still a lot of kinks to be worked out — for example, people still have the wrong number of fingers, and it will occasionally accidentally generate videos of people slamming doors on their own legs and not appearing to notice.

However, that’s not the reason everyone was talking about it. Instead, it’s because of the concerningly hands-off approach that OpenAI appeared to take with copyright. In short, the company seemed to train their model on a bunch of copyrighted content and then told the copyright holders that if they didn’t want to be included, they’d have to tell OpenAI on a case-by-case basis with no blanket opt-out.

This resulted in a lot of videos that copyright holders may not have exactly loved. Turns out that if the Pokemon company wanted to animate Pikachu fighting in World War II, they would have just done that!

Now, it seems there’s been a ton of industry pressure on OpenAI to *not* infringe on everyone’s copyright. So, they’ve already rolled back the program, putting up more content blockers and making users promise that they had the rights to the media they’re creating, even though OpenAI themselves didn’t have the rights to the media they trained it on.

Plus, if you now try to delete your account on Sora, it will *also* delete your ChatGPT account and remove your API access.

Naturally, people who had been using this to violate copyrights in a similar manner to OpenAI themselves are pretty P.O.’d about this turn of events.

“Yea sora is done,” reads one popular post on the Sora subreddit. “Copyright basically took down almost everything cool that you could of done with sora when it first came out, crazy how it lasted like 2 days, I legit can’t even make a SpongeBob generation anymore.”

Maybe this going away is for the best!