In February 2026, a group of almost 40 independent publishers, coordinated by the Independent Publishers Guild (IPG), sent letters of claim to several leading AI developers. The correspondence alleged that copyright protected books, journals, and other literary works may have been used by the AI developers without authorisation to train large language models. These claims are the first coordinated action in the UK by independent publishers in this space and mark a significant escalation in the emerging global trend of copyright owners turning to litigation against AI companies.
This development arrives against the backdrop of legal debate over the lawfulness of using unauthorised copyright protected for content AI training. The most significant UK case to date on this issue, Getty Images (US) Inc v Stability AI Ltd, was widely anticipated to clarify the boundaries of copyright law in the context of AI training. Instead, the judgment left the central question unanswered. Getty abandoned its primary copyright infringement claims during trial after accepting there was no evidence that training had taken place within UK jurisdiction. The court dismissed Getty's remaining secondary infringement arguments, holding that Stability AI's model did not store or reproduce any of Getty's copyrighted images and therefore could not constitute "infringing copies". You can read further, in-depth analysis of the judgment here. The decision is currently subject to appeal, but for the time being it offers rightsholders little comfort and no binding precedent on whether AI training on unlicensed copyright protected work amounts to infringement.
This lack of certainty is compounded by the UK Government's recent retreat from its previously preferred policy approach. In its March 2026 Report on Copyright and Artificial Intelligence, the Government formally abandoned its proposal for a text and data mining exception with an opt-out mechanism, confirming that "the government no longer has a preferred option". With no legislative reform on the horizon and no clear judicial precedent, the legal landscape remains profoundly unsettled, as covered here.
In the absence of definitive answers from either the courts or Parliament, licensing is increasingly emerging as the pragmatic alternative. This is exemplified by the recent inception of the SPUR (Standards for Publisher Usage Rights) coalition between the BBC, Sky News, Financial Times, Guardian and Daily Telegraph which states its mission is to “establish shared technical standards and responsible licensing frameworks” between AI developers and publishers.
In terms of more practical examples, perhaps the most prominent was Disney's three-year licensing agreement with OpenAI, announced in December 2025, which permitted OpenAI's Sora platform to generate content using over 200 Disney, Marvel, Pixar, and Star Wars characters in exchange for a US$1 billion equity investment and the integration of OpenAI's tools across Disney's operations. Although OpenAI subsequently shut down Sora in March 2026, effectively negating the deal, the agreement remains significant as a proof of concept: it demonstrated that AI companies and major rightsholders can reach mutually beneficial commercial terms without waiting for judicial or legislative intervention.
Conversely, the lack of legislative intervention and uncertainty about the Government’s plans may itself be stifling the licensing market. The Publishers Association’s March 2026 report, UK Publishing and the AI Licensing Market, evidences an established and maturing licensing market of a scale and maturity “far beyond previous estimates”, with the number of publishers active in the market expected to nearly double by the end of 2026 and all major UK academic publishers expected to be participating. The report notes that the true value of the market is “many multiples of public estimates”, with publishers entering multi-million pound deals with predominantly US-based technology firms, demonstrating the significant commercial opportunity that legal uncertainty is placing at risk.
This interplay between litigation and licensing is not unique to the United Kingdom. In Germany, the GEMA case concerning AI memorisation of song lyrics was deliberately brought as a test case to provide an impetus for licensing negotiations to take place, illustrating how strategic litigation can serve as a catalyst for commercial agreements rather than an alternative to them. The European position may be further clarified by the pending CJEU referral in Like Company v Google, which held its first hearing on 10 March 2026 and represents the first occasion on which the Court of Justice has directly considered generative AI and copyright.
Copyright owners around the world, from media outlets in Japan and Canada to authors in the United States, are pursuing similar claims against AI developers, reflecting a common global trend. For UK rightsholders, the path forward remains unclear. Litigation may ultimately establish the legal principles that are so sorely needed, but in the interim, licensing frameworks offer the most immediate prospect of progression.
If you have questions about copyright and AI – from enforcement strategy and licensing to navigating the evolving legal landscape – we would be delighted to discuss at INTA 2026 in London. Please contact Adam Glass or Daniel Smith to arrange a meeting.