A US court ruled that Anthropic may proceed to trial on claims it used copyrighted books for model training, rejecting a motion to dismiss on fair-use grounds, but also allowing Anthropic's fair-use defense to advance to trial. The distinction matters structurally: fair-use doctrine is highly fact-specific, and the decision leaves open the possibility that courts will ultimately permit large-scale copyrighted training if vendors can demonstrate transformative use, lack of direct market substitution, or commercial necessity. For financial institutions financing AI vendors, this signals ongoing regulatory uncertainty on training-data IP liability, but does not yet establish that liability is quantifiable or will force retroactive settlement costs.
251 other tracked developments touching the same claims.