Bluesky Considers Proposal for User Consent on AI Data Usage
At the recent SXSW conference in Austin, Jay Graber, the CEO of Bluesky, discussed the company’s efforts to create a framework that gives users control over how their data is utilized for generative artificial intelligence (AI). Although Bluesky, a decentralized social media platform, does not intend to utilize its users’ posts to train its own AI models, the rising demand for training data in AI has prompted them to rethink their policies regarding AI.
Bluesky’s open nature has attracted the attention of other entities looking to use its content for AI training. A notable instance occurred last year when a dataset compiled from one million Bluesky posts was discovered on the Hugging Face platform. This finding raised questions about users’ data rights and the implications of their content being used without explicit consent.
In contrast, Bluesky’s rival platform, X (formerly known as Twitter), has actively incorporated user posts to enhance its own AI chatbot, Grok, through its affiliated company, xAI. Last year, X modified its privacy policy, permitting third parties to utilize user posts for AI training. This policy change coincided with the U.S. elections and heightened the scrutiny on X’s management, which led to many users migrating to Bluesky, contributing to the latter’s impressive user base growth of over 32 million in just two years.
During the SXSW event, Graber emphasized that Bluesky is committed to user agency, explaining that the platform is collaborating with other partners to establish a robust framework for user consent regarding data usage for generative AI. He underscored the importance of allowing users to determine how their content should be utilized, affirming the principle of user choice.
Graber compared this approach to how websites communicate whether they wish to be indexed by search engines. Although search engines can still access content on public sites, the robots.txt file serves as a guideline, honored by numerous search engines regarding data scraping. For Bluesky’s plan to be effective, it would require widespread adoption among users, organizations, and regulatory bodies.
The proposal that Bluesky is currently developing suggests obtaining user consent at either the account or specific post level. Once consent is secured, Bluesky would encourage other companies to adhere to these preferences. Graber noted in his talk that this initiative represents a proactive step towards addressing how AI influences the perception and usage of personal data, indicating a general shift toward more ethical standards in AI deployment.
Overall, Bluesky’s plans signal a growing recognition of the importance of user consent in the age of artificial intelligence, aiming to empower individuals to take control over their own data in an increasingly connected digital landscape.
