OpenAI wants more training data. So it’s building the next X.

When people said they wanted an alternative to Twitter now that it’s been X-ified, this probably isn’t what they meant: ChatGPT, DALL-E, and Sora owner OpenAI is apparently building its own social network to compete with Elon Musk‘s site.

The Verge reports that OpenAI co-founder/CEO Sam Altman has been privately asking for feedback about the project. But he’s also not so privately floated the idea; back in February, when news surfaced that Meta might be looking to make its own ChatGPT competitor, he sniped, “ok fine maybe we’ll do a social app.”

It seemed like a sarcastic comeback at the time, but now–not so much. A person familiar with the matter (but who works at another AI lab, not OpenAI) told The Verge that X integrating Grok into its feed “has made everyone jealous.”

“Especially how people create viral tweets by getting it to say something stupid,” they added.

Like many of X’s updates post-Musk buy, the Grok integration was widely disparaged and caused yet another mass exodus of creatives from the platform, since they didn’t want their art, writing, photos, and videos to be scraped for training or stolen and “remixed” with the click of a button.

And that’s the crux of what’s happening here. Yes, Altman’s beef with Musk and Meta might be some of the inspiration for it to build its own social network. But also, OpenAI is clearly taking stock of its biggest competitors and realizing they have something it doesn’t: vast, constantly updating troves of user-generated content, platforms whose terms of service make everything posted on them takable for training.

Since ChatGPT first appeared on the net, OpenAI and other AI developers have faced scrutiny over where they source the data used to train their large language models, and it’s become more difficult for them to simply take what they want, citing “publicly available” as permission. With platforms like Facebook and X, though, scraping permission can be baked in to basic TOS–so if people want to use those platforms, they are required to consent to pouring a neverending well of fresh meat into the AI grinder.

Now, it’s possible for developers like OpenAI to broker deals with entire platforms–or, in some cases, individual creators–and pay to scrape their data. However, since Meta and X (which Musk recently merged with his AI startup, the creatively named xAI) are competing with OpenAI on artificial intelligence development, they’re obviously not going to give it their training data, leaving OpenAI high and dry for new material.

Unless it can make its own platform.

As The Verge writes, it’s not clear yet whether OpenAI’s envisioned site would be separate from ChatGPT, or if it would be similar to the X/Grok integration. Considering ChatGPT was the most downloaded app on Earth last month, that might be the move OpenAI goes with.

Whether people will actually adopt it as the next big social platform…We’ll see.