The year 2025 proved to be a pivotal time in the ongoing saga of artificial intelligence and copyright law. With over 70 infringement lawsuits already filed by copyright holders against AI companies, the legal landscape has been anything but quiet. We saw significant developments, from crucial court decisions on fair use to substantial settlements that are shaping how generative AI models are trained and deployed.
One of the most talked-about events was the $1.5 billion settlement in the Bartz v. Anthropic case. This massive figure stemmed from Anthropic's use of millions of allegedly pirated books for training its large language model. While a district court initially ruled that Anthropic's training process was "exceedingly" transformative and thus qualified as fair use, the potential for billions in statutory damages for the downloaded works clearly pushed the parties toward a settlement. Announced in September, the agreement saw Anthropic paying approximately $3,000 for each of the nearly half-million books it acquired from pirate libraries. This outcome sent ripples through the AI development community, many of whom are suspected of using similar, piracy-laden datasets. As the Copyright Alliance CEO noted, the settlement demonstrated that AI companies can indeed compensate copyright owners without stifling innovation.
Interestingly, a parallel case, Kadrey v. Meta, also saw a summary judgment order from a Northern California district court just two days after the Bartz decision. Here, authors sued Meta over the use of their books for training. The court also found the use to be "highly transformative" and fair use, but this ruling was notably narrow, largely due to a lack of evidence presented by the plaintiffs' counsel. What makes Kadrey particularly significant, however, is the court's detailed discussion on the potential indirect substitutional impacts on copyright owners' markets. Unlike Bartz, the Kadrey court seemed to grasp the broader implications of generative AI on creative incentives, suggesting a more nuanced approach is needed.
While the Kadrey plaintiffs faced a setback on the fair use question, the case isn't over. It's now proceeding on the issue of whether Meta simultaneously uploaded copyrighted works while downloading them using BitTorrent technology – a process known as seeding. If found liable for distributing pirated content on such a massive scale, Meta could face staggering damages, potentially leading to another significant settlement in 2026. More importantly, the judge in Kadrey has provided a clear roadmap for future copyright holders looking to navigate AI infringement claims.
Beyond these high-profile lawsuits, the latter half of 2025 saw a flurry of other settlements, particularly in the AI music space. These agreements often included not just compensatory payments but also licensing deals, signaling a growing trend of collaboration and formal partnerships between AI developers and copyright holders. A notable example is Universal Music Group's settlement with Udio, announced on October 29th, which included both a legal settlement and a licensing agreement, indicating a path forward for integrating AI music generation with existing copyright frameworks.