Table of Contents
AI training has become a social-network fault line
For years, social platforms have treated user content as the raw material that keeps their products running: posts shape feeds, reports inform moderation, and engagement signals help decide what people see next. Generative AI has made that relationship more contentious. A post is no longer only something shown to an audience; users increasingly worry it may become input for systems that produce new text, images, or other output.
Bluesky, a growing social network, has taken a clear position in that argument. It announced that it will not use user-generated content to train generative AI tools, a declaration that arrived as X, formerly known as Twitter, introduced terms of service allowing public posts to be used for AI model training.
The distinction matters because “AI” is often used as a catch-all term. It can describe ranking systems, spam detection, recommendation engines, moderation tools, and generative models, even though those uses raise different questions about consent, privacy, and creative ownership. Bluesky’s message is not that the service avoids AI entirely. Its promise is narrower, and more meaningful for many users: the platform says their content will not be used to train generative AI.
“We do not use any of your content to train generative AI, and have no intention of doing so.”
Bluesky said many artists and creators are concerned about platforms using their data for AI purposes. That concern is easy to understand. Creators publish work to be read, shared, discussed, and sometimes monetized. Training a generative system on that work can feel like a separate use entirely, particularly when users have little practical ability to track how their material is handled once it enters a platform’s systems.
The company’s wording also offers something users increasingly demand from social networks: a direct explanation of what a policy does and does not cover. Terms of service can be broad by design, and people rarely read them until a policy shift becomes controversial. A plain statement about generative AI training gives Bluesky a sharper public position than vague assurances about improving services.
Related reading: Bluesky’s Rapid Growth — SquaredTech
Moderation and discovery are not the same as generative AI
Bluesky has clarified that it does employ AI for content moderation and its “Discover” feed. Neither of those systems, it says, involves generative AI trained on user content. This is an important qualification rather than a loophole. Large social services need ways to identify harmful material and organize an overwhelming volume of posts. Automated systems are part of that work, whether users like the label “AI” or not.
Content moderation is concerned with identifying or responding to material that may violate rules. A discovery feed, meanwhile, attempts to surface posts that may be relevant or interesting. Those uses can still raise legitimate questions about how a platform makes decisions, what signals it relies on, and whether users can understand or influence the result. But they are not automatically equivalent to using people’s creative output as training data for generative AI.
That separation is where Bluesky’s policy has found appeal among users seeking stronger data protection and ethical AI practices. The platform is making a value judgment: a social network can use automated tools to operate its service without treating every public post as material for generative model development. Whether that position becomes a durable competitive advantage will depend on execution and trust, but it is a notably different pitch from the one emerging at X.
X embraces a broader claim over user data
X’s approach has been more controversial. The platform has outlined how its new policies allow user posts to be analyzed for machine learning and AI model training. Its terms provide broad rights to use user data for improving services, including generative AI applications.
From a platform operator’s perspective, that is an understandable commercial interest. Social networks sit on enormous streams of current language, cultural references, arguments, reactions, and media. Those streams can be valuable for systems intended to understand and generate human communication. From the user’s perspective, though, the same policy can look like a major expansion of what posting publicly means.
Public visibility has never necessarily meant unrestricted reuse. People may post a joke, a drawing, criticism, reporting, or a personal observation because they expect it to be part of a conversation. They may not expect that contribution to help train a generative AI application. The gap between those expectations and a platform’s contractual rights is at the center of the backlash.
X’s policy has drawn criticism particularly from creators and privacy advocates. Elon Musk’s alignment with political figures, including U.S. President-elect Donald Trump, has further polarized opinions about the platform. That political context means the dispute is not only about technical policy. For many users, decisions about where to post are becoming bound up with questions of governance, speech, ownership, and the kind of online public square they want to support.
Related reading: The World’s Most Powerful AI, Trained with Your X — SquaredTech
Growth gives Bluesky an opening, not a guaranteed win
Bluesky has witnessed significant growth following the U.S. presidential election, gaining over 1 million new signups in just 24 hours and reaching 17.14 million users. The surge highlights demand for alternatives as X leans towards more conservative ideologies.
Rapid signups are a signal of dissatisfaction, but they do not automatically settle the competition. Social networks are difficult to leave because their value is tied to the people already there: friends, professional contacts, communities, readers, and sources of information. A privacy or AI policy can prompt people to try a rival service, yet turning that interest into a lasting habit requires active conversations and reasons to return.
That is why Bluesky’s stance has significance beyond a single policy announcement. It gives prospective users an answer to a practical question: what is different here? Product features can be copied. A public commitment not to use user-generated content for generative AI training is harder to reduce to a simple feature comparison because it addresses the relationship between a platform and the people supplying its content.
Meta’s Threads remains the largest competitor to X, with over 275 million monthly active users and 15 million new signups in the past month. That scale is a reminder that Bluesky is not competing in a two-player contest. Threads has the audience advantage, while X retains the weight of its established network. Bluesky’s opportunity is to offer a more distinct set of principles to users who do not see size alone as the deciding factor.
Bluesky’s decision not to exploit user data for AI training could further differentiate it in the social media landscape. The phrasing of platform policies will matter more as generative AI becomes embedded in mainstream products. Users are likely to ask increasingly direct questions: Is my content being used? For what purpose? Can the company explain the difference between moderation, recommendation, and model training without hiding behind technical language?
Bluesky has provided one answer. X has chosen another. The resulting divide is less about whether social networks will use AI than about who gets to set the boundaries around the data that makes those systems possible.
Stay Updated: Tech News – Mobile – Artificial Intelligence
