Social network Bluesky says it won’t train AI on user posts

Social network Bluesky has taken a firm stance against using user-generated content for AI training, distinguishing itself from competitors in an increasingly AI-focused social media landscape.

Key policy announcement: Bluesky has explicitly stated it will not use user content to train generative AI systems, marking a significant departure from other social platforms’ approaches.

• The platform made this declaration just before competitor X implemented new terms of service allowing user content to be used for AI training
• Bluesky emphasized its commitment to protecting artists and creators who have made the platform their home
• The company clarified that while it uses AI for content moderation and its Discover feed, these are not generative AI systems trained on user content

Technical considerations: Despite Bluesky’s stance, the platform’s current technical infrastructure leaves open the possibility of third-party AI training on user data.

• The platform’s robots.txt file doesn’t exclude crawlers from major AI companies like Google and OpenAI
• Spokesperson Emily Liu acknowledged that, similar to regular websites, robots.txt files may not prevent outside companies from crawling the platform
• The company is actively exploring ways to ensure outside organizations respect user consent

Platform growth and challenges: Bluesky is experiencing rapid user expansion while working to maintain platform integrity.

• The platform has gained over three million users in the past week, reaching 17 million total users
• This growth has led to increased spam, scam, and trolling activity
• The company is expanding its moderation team to handle the increased user load

Competitive landscape: The microblogging space is seeing significant shifts as users explore alternatives to X.

• Meta’s Threads has reported over 15 million new signups this month
• Threads is testing custom feeds, a feature already established on Bluesky
• Unlike Bluesky, Meta has admitted to training its AI models on public user posts dating back to 2007

Future implications: While Bluesky’s stance on AI training data sets it apart from competitors, the challenge of preventing third-party AI training of public content remains an industry-wide issue that may require technical and policy solutions beyond individual platform commitments.