Bluesky flags LLM training on unlicensed copyright materials

Bluesky has publicly asserted that most commercially available large language models were developed using copyrighted materials without transparency, permission, or compensation. The statement, posted by Bluesky representative Ruth Booth on October 14, 2026, targets the foundational training practices underlying current AI services. Booth emphasized that this lack of consent applies specifically to models serving as infrastructure for downstream AI applications. The post does not name specific companies but characterizes the issue as systemic across the commercial LLM landscape. This declaration aligns with ongoing global disputes regarding fair use and data scraping in generative AI development. Legal experts note such public statements may signal shifting norms among tech platforms that previously remained neutral on training data provenance. The assertion adds pressure to pending litigation and legislative efforts seeking mandatory disclosure of training datasets. No AI company has yet issued a formal response to this specific claim.