Anthropic has stated that copying protected works as an intermediate step to create non-infringing AI outputs may qualify as fair use, according to their recent response to the Copyright Office.

This position comes amid ongoing legal challenges, including a lawsuit from Universal Music, Concord, and ABKCO alleging copyright infringement. Anthropic maintains that their AI assistant Claude's training process "qualifies as a quintessentially lawful use of materials."

Key points from Anthropic's argument:

• Copying is used only as an intermediate step for statistical analysis
• The process extracts unprotectable elements to create new outputs
• The use is non-expressive and doesn't communicate copyrighted content to users
• The transformation of data for analysis is unrelated to the original work's purpose

ASCAP strongly opposes this interpretation, stating:

• No scenario exists where unauthorized AI training on copyrighted works constitutes fair use
• The use of copyrighted materials for AI models is not transformative
• Each unauthorized use serves a commercial purpose

The debate continues as music industry organizations push for mandatory AI training disclosures in the European Union's AI Act. Meanwhile, AI integration in music platforms continues to expand, with recent launches including Stable Audio and new AI features on Spotify.

This ongoing discussion highlights the growing tension between AI development and copyright protection, with significant implications for both technology companies and content creators.

Related Articles

Previous Articles