Bluesky Will Use User Data To Train AI, But Approval

Bluesky AI training with user data (photo: dock. Bluesky)

JAKARTA Bluesky had become a place for X users to escape in a number of countries because their posts were used as a learning tool for the Grok Artificial Intelligence (AI) model. This policy was controversial.

Many don't like the policy because not a few users upload their artwork on X, from fanatics to short stories in the form of threads. Users X worry that their work will be used to train Grok.

Unlike Bluesky, the platform has always emphasized that it really appreciates artists and creators so that user data is not used to train AI models. However, this statement seems to change in the near future.

Bluesky, which initially used other data to train AI systems, now requires user data as well. At the SXSW conference, Bluesky CEO Jay Graber said they were working on a framework related to user data usage.

Later, this framework will be a feature that can request approval. If users are willing to use all of their posts as an AI training tool, Bluesky will retrieve user data and use it, exactly as X did.

We really believe in user choices, Graber said, citing from TechCrunch on Tuesday, March 11. "It could be similar to how websites determine whether they want their data to be cropped by search engines or not."

Even though Bluesky ultimately trains their AI using user posts, the decision remains with users. Bluesky wants to develop AI in a positive direction in a positive way as well.

"We've been working on it with other people in a field that cares about how AI affects the way we look at the data," Graber said. "I think this is a positive direction."