Bluesky says it has “no intention” of taking user content to train AI tools, ahead of X implementing new terms of service that let it train AI on user content
‘We do not use any of your content to train generative AI, and have no intention of doing so.’
The VergeJay Peters
Context & Ripple Effects
The story captures a widening split over whether platform posts are an internal AI-training asset or remain outside that use. Similar user pushback had already prompted Zoom to clarify limits on training with communications-like customer content, while Reddit was considering tighter access if it could not secure data agreements with AI companies.
Bluesky's position later faced a practical boundary: after a third party published a large dataset from its API, the company said outside organizations were responsible for respecting consent. Its subsequent proposal for user-set AI-training and archiving preferences shows the issue moving from a blanket pledge toward granular controls.
First-order effects
Bluesky can immediately distinguish its service from X by pledging not to use users' content for generative-AI training, making data-use policy part of the platforms' competitive positioning.
X's new terms make its users' posts available for the company's AI training under the stated policy, while Bluesky users receive the opposite internal-use commitment.
Second-order effects
The contrast raises pressure on social platforms to state whether AI training is covered by their terms, especially as earlier disputes pushed companies such as Zoom to clarify their policies.
A platform's own policy does not necessarily govern API-derived copies: Bluesky's later response to a third-party Bluesky post dataset illustrates the gap between internal training promises and external collection.
Third-order effects
If platforms increasingly market consent as a differentiator, AI-data access is likely to shift from broad terms toward more explicit user choices and access controls; Bluesky's later preference proposal is an early sign of that direction.
The durable fault line is likely to be governance over public data: platform assurances can constrain first-party use, but meaningful control also depends on how APIs, researchers, and other downstream collectors handle content.
The trend: Social platforms are turning AI-training rights over user content into a product-policy and data-governance battleground.
A number of artists and creators have made their home on Bluesky, and we hear their concerns with other platforms training on their data. We do not use any of your content to train generative AI, and have no intention of doing so.
Bluesky uses AI internally to assist in content moderation, which helps us triage posts and shield human moderators from harmful content. We also use AI in the Discover algorithmic feed to serve you posts that we think you'd like. — None of these are Gen AI systems trained on …
The explicit promise is appreciated. — However, ownership can change and such intentions may shift. Users may not be able to act in time to protect their content. — What happened at #Google is an example of this. Once intent on “do no evil” it has become as bad or worse tha…
This is important. I know this is kinda like preaching to the choir on this site but the evils and harm done by genAI is massive, and it doesn't just affect artists. It affects everyone negatively. Keep on ceaselessly fighting against this blight. [embedded post]
That's good to hear. While my twitter account's still up, I've been using Bluesky as my primary platform lately and I've been digging it. [embedded post]
This is especially good to hear knowing that today is the day Twitter puts their mandatory AI training into effect (alongside the new and unimproved block feature). — Happy to be here and I hope for these blue skies to stay nice and clear for everyone — 💙🦋 [embedded post]
Kinda goes without saying, but you should continue to use things like Glaze, Nightshade, or add AI disruptors on your art wherever you post. Bluesky is just giving the guarantee that they *specifically* will not use art posted here for AI training (which is a very good thing, an…
this is genuinely wonderful to hear, given how many social media platforms (and even art ones) will fold at the mention of ai, it's refreshing to see a big upcoming platform like this is against it 🙏 — i think it'd be wonderful if they could ensure our work can't be fed into ot…
I commend Bluesky for making this statement and I truly hope their stance never changes. We creators are tired of leaping around from platform to platform after having our work used to train their AI. [embedded post]
In Neal Stephenson's book “Fall, or Dodge in Hell”, he describes a type of AI bot called APES. APES appear as humans, but post nonsense in every direction. Good, bad, mild, extreme. They dilute the old internet, becoming meaningless. Their purpose: Destroy it so a new net can…
Also, [upside-down peace sign dog picture] — Instagram, Threads, and X all train for their 1st-party gen AIs so scramble outta there if you can afford to. [embedded post]
Personally, even though I don't consider myself a prolific creator or artist (more like a hobbyist), it is relieving to see this made clear for everyone. [embedded post]
I am pretty new around here and have had a great time so far. — And then Bluesky does something shocking and makes user centric, transparent and smart decisions regarding their business practices. — I could really get used to such a company. It's refreshingly awesome. [embe…
As a developer: Bluesky is like the “good old” Twitter, before the API got taken away/locked down Eg you can access the “firehose!” Without having to pay $42K/month or more One of many reasons it's becoming popular with devs (and will continue to do so) Love it! [image]