Lawsuits could spell trouble for AI

The rise of AI has been both awe-inspiring and unsettling to experience, mostly because it’s unfolded so seamlessly, so quickly, into our everyday lives.

Since our 2023 report on genAI and journalism, advances in AI technology have brought us closer than ever to mimicking human-like interactions with computers. Complex reasoning models and agentic AI capable of autonomous decision-making are becoming more prominent. Multimodal systems that simultaneously process text, images, audio, video, and code within a single prompt, along with intelligent assistants and chatbots, are par for the course. AI tools are now embedded in search engines, productivity software, and consumer devices, making AI almost as easy to access as the internet itself.

In short, AI is everywhere, with no signs of disappearing.

What may slow AI’s seemingly inexorable expansion is the rising wave of copyright infringement lawsuits, many of them brought by news organisations. These include The New York Times’ case against OpenAI and Microsoft, as well as actions by Dow Jones & Company and numerous others in the US, Canada, India, and Europe. Large-scale AI models, such as LLMs like GPT, Claude, and Gemini, rely on vast and highly varied datasets to produce diverse, contextually relevant outputs across topics and user needs. High-quality material, such as professionally produced news content, makes it valuable for training and for enhancing user experience.

At issue are closely linked questions about how AI companies acquire and use their training data. Central to this is whether copyright exceptions – fair use in the US and fair dealing in Australia – permit AI developers to use copyrighted works for training without permission, and whether producing substantial reproductions of those works constitutes infringement. Neither Australian nor US copyright law directly addresses AI, leaving these issues to be judicially tested. For the news industry, the stakes are high: beyond concerns over attribution and compensation, there is the risk of market harm, where uncredited replication of journalism diverts audiences and revenue away from publishers and towards AI platforms.

This month, major Japanese newspaper Yomiuri Shimbun became the latest news organisation to sue, filing a case against Perplexity for allegedly scraping thousands of its articles, which were later reproduced in response to user queries. Under Japanese copyright law, third parties may use copyrighted works for training purposes provided the use is not for enjoyment or does not enable others to enjoy the works, a condition that would generally exclude wholesale reproduction.

As Monica has mentioned, Australia’s Productivity Commission is floating the idea of introducing a similar text and data mining exception to our own Copyright Act, much to the chagrin of local authors and creators whose works have already been scooped up, without permission or payment, in AI training data.

While most cases are still before the courts, two California judgments from June 2025 indicate that any determination will ultimately depend on the facts. In Anthropic and Meta AI,  both brought by book authors, the courts found in favour of the developers’ argument that the training process constituted fair use because it was transformative (i.e., it added something new to the original works). In Meta, the court’s finding turned on the authors’ failure to present evidence that Meta’s use of their works caused market harm. In Anthropic, the decision was narrowly applied only to those books which the developer had legally obtained – not to the some seven million works it had pirated, an issue the court will hear later. Significantly, in that issue, the court has decided that the three authors suing Anthropic could represent all writers nationwide whose works had been pirated by the developer.

However, AI industry groups and some author advocates are backing Anthropic’s bid to overturn that decision,  in what’s being called the largest class action ever certified. Anthropic argues that it could financially devastate not just itself – with damages reaching into the billions if required to account for all claimants – but the broader AI sector. They argue that the court rushed its decision, preventing rigorous analysis of class membership and ownership rights, overlooking potential orphan works with no identifiable owner, split rights and defunct publishers. The sheer size of the case could very well be overwhelmed with administrative complexity, force a rushed settlement, and prevent proper examination of the legal issues – which this field sorely needs.

With policy changes around copyright law and AI currently under review in Australia, these judgments will nevertheless be pivotal in shaping the future of AI given that most of the major developers are US-based. They will help set legal boundaries on how copyrighted works may be used in training and in determining the extent of liability. Without access to updated, quality training data, AI machines will suffer.

To avoid this – and further litigation – AI developers are entering licensing deals with large news publishers in Europe, North America and some in Australia. While this raises questions around how AI developers plan on using copyrighted works outside these regions and companies, it is a step in the right direction to properly acknowledging the importance of journalistic works.

References 

Japan’s largest newspaper, Yomiuri Shimbun, sues AI startup Perplexity for copyright violations | Nieman Lab. https://www.niemanlab.org/2025/08/japans-largest-newspaper-yomiuri-shimbun-sues-perplexity-for-copyright-violations/  

Watering down Australia’s AI copyright laws would sacrifice writers’ livelihoods to “brogrammers.” Spicer, T. (2025, August 12) | The Guardian. https://www.theguardian.com/commentisfree/2025/aug/12/australia-ai-weakening-copyright-act-laws-productivity-commission

The Productivity Commission is floating AI copyright exemptions – with worrying implications for Australian authors and publishers. ‌Grundy, A. (2025, August 6). | The Conversation. https://theconversation.com/the-productivity-commission-is-floating-ai-copyright-exemptions-with-worrying-implications-for-australian-authors-and-publishers

‌‌Case Tracker: Artificial Intelligence, Copyrights and Class Actions. | BakerHostetler. https://www.bakerlaw.com/services/artificial-intelligence-ai/case-tracker-artificial-intelligence-copyrights-and-class-actions/  

What comes next for AI copyright lawsuits? Heaven, W. D. (2025, July 1) | MIT Technology Review. https://www.technologyreview.com/2025/07/01/1119486/ai-copyright-meta-anthropic/  

US authors suing Anthropic can band together in copyright class action, judge rules. Brittain, B. (2025, July 18) | Reuters. https://www.reuters.com/legal/government/us-authors-suing-anthropic-can-band-together-copyright-class-action-judge-rules-2025-07-17/  

AI industry horrified to face largest copyright class action ever certified. Belanger, A. (2025, August 9) | Ars Technica. https://arstechnica.com/tech-policy/2025/08/ai-industry-horrified-to-face-largest-copyright-class-action-ever-certified/ 
‌ 
Platforms and publishers: AI partnership tracker. ‌Pete Brown (2025) | Quarto https://petebrown.quarto.pub/pnp-ai-partnerships/