The UK publishing industry is intensifying its calls for government intervention, urging ministers to compel artificial intelligence (AI) companies to license copyrighted content. This renewed pressure follows a significant development in the United States, where a federal court approved a $1.5 billion (approximately £1.1 billion) class action settlement with AI developer Anthropic.
The settlement addresses Anthropic's storage and use of millions of pirated books to train its AI models, marking one of the first major AI copyright disputes to result in substantial compensation for authors and publishers. Dan Conway, chief executive of the Publishers Association, stated that this agreement should serve as a stark warning to AI developers operating in Britain. He argues that the UK government must now implement robust regulations to ensure technology firms financially compensate creators for the material utilised in training their AI.
Bloomsbury Publishing confirmed on Tuesday that it is a beneficiary of the settlement, with more than 14,000 of its titles included. The publisher anticipates receiving payments later this year, with eligible works valued at roughly $3,000 per title, before legal fees, to be equally divided between authors and publishers. This resolution comes after a previous landmark ruling found Anthropic's use of copyrighted books for training to be 'fair use' under US law, though its retention of pirated books fell outside this protection, leading to the settlement to avoid a potentially larger trial.
The UK government previously abandoned proposals that would have allowed AI companies to train models on copyrighted works unless creators explicitly opted out. Instead, ministers have committed to supporting licensing markets and exploring mandatory measures for AI developers. In May, Baroness Keely, chair of the House of Lords Communications and Digital Committee, welcomed the shift away from the opt-out model but cautioned that voluntary transparency alone would be insufficient. She stressed that only a mandatory framework could establish the necessary level playing field for responsible data practices.
Mr Conway highlighted that the US settlement specifically relates to the past use of pirated content and does not definitively establish a broader requirement for AI developers to seek permission and pay for all copyright-protected works in AI training. He suggested that further litigation will be needed to fully clarify whether AI firms require licences before using such material. Jonathan Quinn, a partner specialising in AI litigation, underscored that courts are only just beginning to grapple with the novel legal questions posed by generative AI, indicating that many fundamental answers remain elusive.