\n\n
Key developments:
- Major book publishers are secretly employing LLMs such as Claude and ChatGPT for tasks ranging from publicity copy to cover art, reflecting a growing internal reliance on AI despite public opposition.
- The Third Circuit Court of Appeals affirmed that Westlaw headnotes are copyrightable and that Ross Intelligence’s use of those headnotes to train its AI legal‑research tool was not fair use, marking the first appellate decision directly addressing AI‑training copyright issues.
- The ruling underscores that AI developers cannot assume a blanket fair‑use defense for training data, even as other high‑profile disputes loom, including a pending case against Anthropic for allegedly using copyrighted song lyrics.
Why it matters:
- The decision sets a legal precedent that AI training is subject to the ordinary fair‑use analysis, potentially limiting how companies scrape copyrighted material for model development.
- Publishers’ covert AI usage raises ethical and contractual questions about transparency with authors and the ownership of AI‑generated promotional material.
- Ongoing lawsuits, such as the upcoming Concord Music Group case against Anthropic, indicate that the copyright battle over AI training data is far from settled and will shape industry practices.
What to watch next:
- Further appellate opinions on AI training and fair use, especially in the music and publishing sectors.
- Legislative proposals that could require AI companies to disclose training data sources.
- Corporate responses, including licensing agreements or shifts to publicly available datasets, as the legal landscape tightens around AI model development. [[1, 2, 3]]