From Combat to Collaboration? How AI Model Companies are Licensing Content
The need for content has never been greater. Particularly in the generative AI model space, as AI companies continue to compete with one another for market share. However, content comes at a price. Since generative AI models require vast amounts of training data, portions of the content used to train AI models are protected under copyright law. Large-scale battles have erupted over the use of unlicensed content to train models. As of December 5, 2025, 65 copyright lawsuits have been filed against AI companies in the US[i]. As of the time of this blog post, the latest lawsuit was filed by The New York Times (“NYT”) against Perplexity[ii]. Perplexity has been sued many times – by The Chicago Tribune, Dow Jones (The Wall Street Journal), the New York Post, and other publications.
As these 65 copyright lawsuits progress, the resulting decisions or settlements will ultimately determine the path forward for AI companies and content owners, as well as how they interact with one another. This blog post addresses how the marketplace and business models for content are evolving, surveys the latest trends, and provides a few predictions.
Players and Tactics
At one end of the spectrum, we have the early first-mover AI companies getting their AI models to the market as fast as possible, without licensing content (e.g., OpenAI, Anthropic, and others). On the opposite end of the spectrum are AI companies playing a longer game – entering partnerships with content owners first and training their models with only licensed content (e.g., ProRata, Bria, PipLabs, and others). In the middle of the spectrum, there are AI companies whose actions represent a mix of the two approaches.
First Mover Approach
This approach is best characterized by companies pushing the boundaries of copyright law and the rights of content owners. A recent example of this is OpenAI’s release of Sora 2, its flagship video generation tool. Sora 2 allowed users to create videos featuring copyrighted material. Upon Sora 2’s release, OpenAI required content owners to specifically opt out of having their protected content appear in the users’ outputs. OpenAI has since retracted this approach due to an uproar from prominent companies and individuals in the entertainment community[iii]. Despite its retraction, the strategy remains clear. By effectively placing the burden on content owners to opt out of Sora 2’s use of their content, OpenAI was willing to push the boundaries of copyright law to get to market with Sora as quickly as possible[iv].
Long Game Approach
Companies at this end of the spectrum are taking the opposite approach by structuring their business around partnerships and building long-term relationships with content owners. ProRata is an instructive example. ProRata develops an AI-powered search, advertising, and attribution platform called Gist that uses AI models trained with licensed content.[v] ProRata’s technology analyzes AI model output and determines contributing content to calculate proportional compensation.[vi] It achieves this through an approach that attributes the sources of the content that contributed to the output, enabling copyright holders to be compensated on a per-use basis.[vii] ProRata takes a longer-term approach, focusing on partnerships with content owners first rather than by getting to the market as soon as possible and starting with an adversarial relationship with the content owners.
Middle Ground
Between the two ends of the spectrum are companies that are pushing the boundaries of copyright law, while simultaneously building partnerships with content owners. Perplexity, the company mentioned earlier, is a good example of this. Perplexity is an AI-powered search engine that competes with other players in the market, such as OpenAI’s GPT models and Anthropic’s Claude, among others. In October, Perplexity and Getty Images announced a licensing deal, where Getty will provide images from its vast library for Perplexity to display across its various tools.[viii] In return, Perplexity will attribute Getty Images by including image credits and source links so that users can license the images directly from Getty.[ix] Simultaneously, Perplexity has made it clear that the company will continue to fight copyright lawsuits and cement itself as a key player in the AI race. Specifically, AI’s legal chief stated that the company won’t be “bullied” by the influx of copyright lawsuits against AI companies and that they may need to “push the boundaries” to stand up for their “rights.”[x] Given the NYT case filed last week against Perplexity, along with the number of other pending cases against the company, it appears that they are prepared to go to battle with content owners and, in some cases, enter into licensing agreements.
Shifting Landscape and Predictions
Amidst the ongoing litigation and legal uncertainty surrounding copyright protection as it pertains to generative AI, a few AI companies are reevaluating their relationships with content owners and increasingly leaning towards licensing. Perplexity, although it has a strong stance on pushing copyright boundaries, is not the only one making a strategic choice to establish licensing partnerships. AI music generator Udio has recently signed a deal with Universal Music Group, in which the pair agreed to launch a new platform trained on licensed songs.[xi] This partnership between the two companies follows a dispute in which Universal accused Udio of copyright infringement. The settlement is illustrative because it demonstrates that while content owners want to protect copyrighted works, they also acknowledge that AI is transforming the creative processes and business models for content.[xii] At the same time, AI companies are realizing that the “first mover” approach, training their models on unlicensed content to get to market quickly, may not be the most prudent strategy considering the mounting legal risks.
Hence, while some companies continue to push the limits of copyright law, we are witnessing other players adopting a more long-term approach – adjusting their business models to focus on partnerships and sharing upside with content owners. Indeed, licensing partnerships may become standard to manage prospective legal harm and uncertainty while maintaining sustainable business operations. This shift is especially notable considering a key battle on appeal at the Third Circuit – Thomson Reuters v. Ross Intelligence, Inc. As the first copyright infringement case between an AI company and a content owner to reach the appellate stage, the forthcoming decision will serve as a key guidepost for how courts apply the copyright doctrine of “fair use” in the era of generative AI. The Third Circuit’s decision may ultimately influence the next stage in this ever-shifting landscape. To address these legal risks and to encourage partnerships, there is a need for technology platforms that help to enable a marketplace for content.
Stay tuned for part 2 of this blog post, where we explore the Third Circuit case between Thomson Reuters and Ross Intelligence, Inc., and discuss how its outcome may impact the generative AI industry.
By Bob Steinberg, Ben Resnick
Bob Steinberg is the Founder of Generative IQ® LLC, a venture fund that provides capital to early-stage IP rich AI start-ups. He has been protecting and litigating IP rights, working with technology entities and entrepreneurs navigating the IP landscape and monetizing blocking rights for over 30 years.
Ben Resnick is a J.D. Candidate 2027 at Southwestern Law School, where he serves as a Student Editor on the Journal of International Media & Entertainment Law. He graduated from California Polytechnic State University, San Luis Obispo, in 2020 with a degree in Business Administration and a concentration in Financial Management. Ben enjoys playing tennis, hiking, surfing, and reading. He also loves to support the local LA sports teams and explore new restaurants and neighborhoods around LA.
DISCLAIMER: The opinions expressed in this blog post are not provided as legal advice.
