Though I might not be the best reader of legalize, I'm going to try my best to determine what Scratch plans to do with your user content in regards to AI. So, the long and short of it, yes Scratch *can* use any user content to train any AI model. This can include, your sprites (art), sounds, code, or text you put in projects. HOWEVER, that doesn't mean they're going to be using it to undermine the users. Let me go through everything I've found regarding Scratch's position on AI. First, the Terms of Service (ToS) (I didn't actually read through the whole thing so I might have missed something): https://mitscratch.freshdesk.com/en/support/solutions/articles/4000219182-scratch-terms-of-service Now you might have already looked at it or at least seen somebody point out the popular line, "By Posting User Content to or via the Service, ... you hereby grant: ... [us] to otherwise improve our products and services and create new products and services, including through the training of AI models." But there are other mentions of AI in their ToS and they seem to paint a more ethical picture. For reference, they define "Output" as AI-generated content. Under section 11 Prohibited Conduct they say, "BY USING THE SERVICE, YOU AGREE NOT TO: ... 13 automatically or programmatically extract any data of the Service, including Output; 14 represent that Output was human-generated when it was not; or 15 use the Service or any Output to develop or train any AI or machine learning model." 11.13 says you can't extract Output or other data, 11.14 says you can't misrepresent AI content as your own, and 11.15 says you can't use Output or any other user data to train other AI's, which prevents other companies from using our content to train AI. Only Scratch can do that. In other words, Scratch wants to protect our data from being scraped by the likes of OpenAI, Google, Anthropic, eta to train their AI systems. They've built it in where they can't do that without facing legal action. Second, a blog post explaining their AI philosophy: https://medium.com/scratchteam-blog/introducing-creative-ai-at-scratch-bcdbd6c3efc9 Here, they describe how they plan to use AI in an ethical manner. It's a three-minute read and it's not written in any hard-to-read state so feel free to read it. But the TL-DR, if you feel you need it, is this: Scratch plans to protect the data of kids on this platform, they'll be transparent about how the AI is designed, and will leave the creative freedom up to the users. What I can infer (aka unconfirmed but likely true): If you're concerned about environmental concerns, good for you! You must know, though, that Scratch is still a non-profit foundation. They wouldn't have the resources or money to pay for billions of dollars' worth of utility for a warehouse-sized server farm. Final thoughts: I feel your concern about Scratch and their introduction to AI products, but I'm somewhat optimistic. I don't agree with having Scratch introduce AI, even with all these safeguards. But at least I know they're trying to adapt to a changing landscape while keeping their values the same. While yes I guess it can introduce kids to AI in a healthy way, I don't want them to encourage it since what the other AI companies are trying to do is manipulate consumers, not just replacing artists. And teaching kids that AI is good, is not good in my opinion. Conclusion: Yes, Scratch can use your user content to train their AI and yes they're probably doing it now. But at least they're also protecting you from the other companies. One AI training is better than 50. Besides, they're trying to keep it ethical and wouldn't be able to destroy the environment without going net negative. Do I support it? No. Can I accept it? Yeah sure whatever... Don't villainize the non-villains.
