Bluesky Won’t Use User Posts to Train Its Generative AI Models

Recently, Bluesky declared that it does not use user data to build its generative artificial intelligence (AI) models.

In a post on Friday, the social media platform Bluesky stated that it has “no intention” of using user material to train generative AI tools. On the same day that X’s new terms of service, which outline how it can utilize user text and other data to train its generative AI tools, went into effect, it issued the remark.

The generative artificial intelligence (AI) models that Bluesky uses are not trained on user data, the company recently declared. Along with highlighting the areas in which it employs AI tools, the social media platform asserted that none of the models had been trained on user content, either private or public. Following concerns expressed by a number of creators and users regarding the platform’s AI privacy policy, the message was made public. Notably, one million people joined Bluesky in a single day last week, and the platform recently surpassed the 17 million registered users milestone.

In a statement, Bluesky states, “We hear the concerns of several artists and creators who have established themselves on Bluesky regarding other platforms using their data for training.” “We do not and do not intend to use any of your content to train generative AI.”

Bluesky Says It Does Not Train AI on User Posts and Bluesky stated further that their position on AI and user data in a post on the platform. Following concerns expressed by a number of artists and producers on the site on the company’s AI policies, the post stated, “We do not use any of your content to train generative AI, and have no intention of doing so.”

Bluesky also enumerated the domains in which it employs generative AI tools in a different post. As is customary for social media platforms, the company uses AI internally to support its content control system. Furthermore, it employs AI in its Discover algorithmic feed, which makes post recommendations to users based on their platform activity.

Third-party companies can still crawl the site and scrape the data to train their models, according to The Verge, even though the corporation may not be collecting user data to build AI models. “Bluesky’s robots.txt files do not prevent outside companies from crawling its website for data,” company representative Emily Liu told the outlet.

“This is similar to how robots.txt files don’t always stop outside organizations from indexing such websites. We are actively debating how to accomplish this within the team, but we would like to contribute to ensuring that external organizations respect user consent.

The representative did, however, point out that the matter is presently being discussed within the team and that Bluesky is working to determine ways to guarantee that other organizations respect user consent on the platform.

AI is used to “assist in content moderation, which helps us triage posts and shield human moderators from harmful content,” according to a different article from Bluesky. The automated Discover feed also makes use of it. According to Bluesky, “none of these are Gen AI systems trained on user content.” Additionally, the business directs you to a page that contains its community guidelines, terms of service, and other policy documents.

The platform’s safety account indicates that Bluesky has added three million users in the last week. In order to handle the increased volume of “spam, scam, and trolling activity,” the company says it is expanding its moderation staff.

Interestingly, Bluesky said on Sunday that one million new people had signed up for the social media site in a single day. A statistics tracker reports that Bluesky has surpassed 17 million subscribers as users look for microblogging sites other than X. One of its main rivals, Meta’s Threads, isn’t waiting. The platform has already crossed 15 million signups this month, according to Threads CEO Adam Mosseri, who stated yesterday. Today, Meta announced that it is testing personalized feeds, a feature that Bluesky is already well-known for. However, Meta has admitted that, in contrast to Bluesky, it has used practically everything you have publicly posted since 2007 to build its AI models.