Speaking at the SXSW conference in Austin on Monday, Bluesky CEO Jay Graber says the social netlabor has been laboring on a structurelabor for employr consent over how they want their data to be employd for generative AI.
The accessible nature of Bluesky’s social netlabor has already permited others to train their AI systems on employrs’ satisfyed, as was uncovered last year when 404 Media came atraverse a dataset built from 1 million Bluesky posts presented on Hugging Face.
Bluesky competitor X, unkindwhile, is feeding employrs’ posts into sister company xAI to help train its AI chatbot Grok. Last descfinish, it alterd its privacy policy to permit third parties to train their AI on employrs’ X posts, as well. The transfer, pursueed by the U.S. elections that liftd X owner Elon Musk’s status wislim the Trump administration, helped fuel another exodus of employrs from X to Bluesky.
As a result, Bluesky’s uncover source, decentralized X alternative has grown to over 32 million employrs in equitable two years’ time.
However, the demand for AI training data unkinds the novel social netlabor has to slimk about its AI policy, even though it doesn’t arrange to train its own AI systems on employrs’ posts.
Speaking at SXSW, Graber elucidateed that the company has included with partners to grow a structurelabor for employr consent over how they would want their data to be employd — or not employd — for generative AI.
“We repartner consent in employr choice,” Graber shelp, saying that employrs would be able to expound how they want their Bluesky satisfyed to be employd.
“It could be someslimg analogous to how websites expound whether they want to be scviolationd by search engines or not,” she persistd.
“Search engines can still scviolation websites, whether or not you have this, becaemploy websites are uncover on the accessible Internet. But in ambiguous, this robots.txt file gets esteemed by a lot of search engines,” she shelp. “So you need someslimg to be expansively adselected and to have employrs and companies and regulators to go with this structurelabor. But I slimk it’s someslimg that could labor here.”
The proposal, which is currently on GitHub, would include getting employr consent at the account level or even at the post level, then ask other companies to esteem that setting.
“We’ve been laboring on it with other people in the space worryed about how AI is impacting how we see our data,” Graber includeed. “I slimk it’s a selectimistic honestion to get.”
