Bluesky says it's up to outside organizations to respect user consent, after a Hugging Face employee posted a 1M-post dataset from Bluesky's API for ML research
Update: Following the publication of this article on Tuesday evening, van Strien removed the dataset.
It also arrived as Bluesky was consulting on EU compliance, making data-use controls part of a broader question of how the service turns user expectations into enforceable platform rules.
First-order effects
The removal ends public access to this particular 1-million-post research dataset, while leaving Bluesky's position clear: outside organizations, rather than the platform, must honor user consent.
Users and API-based researchers face an immediate mismatch between content being technically accessible and its authors' expectations about machine-learning use.
Second-order effects
Researchers and dataset hosts may face greater pressure to document consent and provenance before publishing social-data corpora, especially where platform APIs expose public posts.
Bluesky has a practical incentive to make its later data-use preference proposal legible to API users; a preference that third parties cannot reliably detect or follow does little to resolve the dispute.
Third-order effects
The episode points to a public-data permission boundary in which API access alone is increasingly insufficient as a social license for AI reuse.
If platforms adopt user-level data-use signals, the competitive question shifts from whether content is public to whether consent can travel with content across APIs, archives, and dataset repositories.
The trend: Social platforms are moving from broad promises about their own AI training toward consent architecture that must govern third-party access and downstream reuse.
@danielvanstrien.bsky.social
Daniel van Strien on bluesky
I've removed the Bluesky data from the repo. While I wanted to support tool development for the platform, I recognize this approach violated principles of transparency and consent in data collection. I apologize for this mistake. [embedded post]
@bsky.app
@bsky.app on bluesky
Brief update on our ongoing efforts to allow users to specify consent (or not) for AI training: 🧵
@danielvanstrien.bsky.social
Daniel van Strien on bluesky
First dataset for the new @huggingface.bsky.social @bsky.app community organisation: one-million-bluesky-posts 🦋 — 📊 1M public posts from Bluesky's firehose API — 🔍 Includes text, metadata, and language predictions — 🔬 Perfect to experiment with using ML for Bluesky 🤗 …
@pikkyu.bsky.social
@pikkyu.bsky.social on bluesky
Due to people already scraping posts here in BSky for training 🅰️ℹ️ garbage. I feel like I need to remind fellow artists who don't already do it to PLEASE protect your artworks from being scraped by using Glaze and/or Nightshade. (1/11🧵) …
@drew.battenhall.com
Drew Benvie on bluesky
To AI or not to AI. That is the question and a very important part of Bluesky's growth: the company pledges to respect user privacy, but also provides complete interoperability of its system & people's content to make it customisable for users. So one user made this. — www.40…
@bsky.app
@bsky.app on bluesky
We recently shared Bluesky's stance on user data and AI training, which has not changed. Bluesky will not train generative AI on user data. bsky.app/profile/bsky... [embedded post]
@bsky.app
@bsky.app on bluesky
Bluesky is an open and public social network, much like websites on the Internet itself. Websites can specify whether they consent to outside companies crawling their data with a robots.txt file, and we're investigating a similar practice here.
@bsky.app
@bsky.app on bluesky
For example, this might look like a setting that allows Bluesky users to specify whether they consent to outside developers using their content in AI training datasets — Bluesky won't be able to enforce this consent outside of our systems. It will be up to outside developers t…