Skip to content

Bluesky Commits to User Privacy: No Generative AI Training on User Posts

    Bluesky Commits to User Privacy No Generative AI Training on User Posts

    Though X’s new privacy policy allows their Grok large language model to be trained on user posts, rival social media app Bluesky revealed no plans to follow suit in an update post released Friday–the day that X’s revised terms went into effect.

    Bluesky has seen significant interest among users and creators following recent changes at X and an increase in leftist users; however, 404 Media notes that its data can still be scraped off the open Internet.

    X’s New Terms of Service

    Some users are opting-out of X because of its new terms of service policy, which took effect November 15. Under this new arrangement, the company was granted a non-exclusive, royalty-free license to utilize content uploaded by users for training and developing AI models including generative AI.

    This has caused concern among artists and others on the platform who fear their work is being used to train computers that could one day replace them, leading some users to delete photos from their feeds altogether.

    X’s new terms of service also includes a provision that requires users to agree to final and binding arbitration in case of disputes instead of going to court, which is an increasingly common provision among tech platforms as an efficient method for dispute resolution.

    However, the clause also stipulates that users agree not to participate in class action lawsuits regarding larger collections of public posts – something digital researchers who rely on such posts may find advantageous for conducting their research.

    Meta’s Threads

    Bluesky stands out as a decentralized social media platform which has experienced remarkable growth since the US election by not incorporating AI development clauses in their terms of service agreements, unlike its peers X and Meta. Bluesky’s decision not to do so highlights an increasing concern regarding data privacy and AI ethics.

    Bluesky plans to develop generative AI models without using users’ posts, and will inform its community when this feature becomes available. Meanwhile, Bluesky continues to enhance features currently offered to its community members.

    One of the updates includes a revamped method for searching content. By providing users with more personalized search experiences that align with its goal of recreating early internet ethos, the company aims to deliver an experience more tailored towards individual user searches.

    Hugging Face’s Scrape

    After the new terms of service from X, Bluesky–which recently emerged as a decentralized alternative–is under increased pressure to clarify its AI data usage policies. Bluesky’s vague statements allowing third parties access users’ posts for various purposes including training AI models have caused alarm among those who assumed that having more transparency and protection provided by decentralizing platforms.

    According to 404 Media, one million public Bluesky posts with identifiable data were culled and submitted for machine learning research by AI company Hugging Face using Bluesky’s Firehose API which provides third-party developers access to user posts and associated data.

    Hugging Face utilizes its library of datasets to assist users in building their own AI systems, while it also hosts cutting-edge pre-trained machine learning models for NLP and other tasks, such as the popular BERT and GPT models. Should commercialized models continue, however, these libraries could become outdated, dissuading users from joining Hugging Face as they would eventually become less attractive options for AI research.

    Bluesky’s Policy

    Bluesky’s declaration comes as a relief to users concerned with how their data is being exploited. While anyone is free to collect public information from social networks such as X and Meta’s Threads for machine learning training (as demonstrated by Daniel van Strien’s release of over one million public Bluesky posts via Hugging Face), Bluesky’s clear statement marks an important step forward.

    Decentralization ethos is also crucial when it comes to building trust among users and safeguarding privacy. Many users signed up for platforms like X, Threads and Bluesky with the expectation that decentralized practices would mean improved data handling practices and transparency compared with those offered by centralized platforms; vague policies allowing loopholes can damage user trust and undermine their promises of decentralized promise.

    Leave a Reply

    Your email address will not be published. Required fields are marked *