Judge: Anthropic’s $1.5B settlement is being shoved “down the throat of authors” (arstechnica.com)

A proposed $1.5 billion settlement between AI company Anthropic and authors over widespread unauthorized use of copyrighted books to train its models has been strongly criticized by US District Judge William Alsup. The lawsuit, which involves up to 7 million authors whose works were allegedly pirated, could have led to damages exceeding $1 trillion, but the settlement covers fewer than 500,000 works and represents a fraction of Anthropic's current $183 billion valuation. Alsup condemned the deal as rushed and incomplete, warning it risks being forced “down the throat of authors” without resolving crucial details like how payments will be allocated or disputes managed. The judge rejected preliminary approval of the settlement, highlighting unresolved issues including the exact list of works, notification processes, claim forms, and dispute resolution mechanisms. He emphasized the need for clear instructions requiring all copyright holders of a work to opt in for coverage, with any dissent potentially excluding that work, and insisted that ownership disputes be handled in state court. Alsup’s decision underscores the complexities and high stakes involved in addressing copyright infringement in AI training—signaling a demand for more rigorous standards to ensure fair compensation and clarity in future agreements involving large-scale use of copyrighted material in AI development.