AI copyright disputes spiked in 2025, and the cases that will define how creative industries use AI for the next decade are already in litigation. The New York Times sued both OpenAI and Perplexity AI. Getty Images' UK case against Stability AI largely failed. UMG settled with Udio and announced a licensing deal. Each outcome reshapes the legal framework for AI-generated content, talent agreements, and content licensing in different ways — and managing partners advising media and entertainment clients need to understand all of them.

California's digital replica law took effect January 1, 2025, prohibiting contracts that create digital replicas without specific protections — and talent agreements across the industry are being rewritten to account for AI. The question isn't whether AI belongs in entertainment anymore. It's who owns what it creates, who controls how it uses existing works, and how talent gets compensated when algorithms replace performances.

The major AI copyright cases fall into three categories, and each will produce different legal standards. Training data cases: The New York Times sued OpenAI in late 2023, alleging millions of copyrighted articles were used to train AI models without consent. In December 2025, the Times filed a second federal lawsuit against Perplexity AI for the same basic allegation. These cases will determine whether using copyrighted material to train AI models constitutes fair use — and the answer will affect every content company in the world. Output infringement cases: Getty Images claimed Stability AI used Getty material to train its image generator, but a UK court largely rejected Getty's claims in November 2025 after Getty couldn't secure evidence about Stability's training process. The evidentiary challenges in these cases are significant — proving what training data an AI model used requires access to proprietary systems. Settlement and licensing models: UMG settled with Udio and announced plans to license its music catalog to the platform, with additional guardrails and a new jointly developed service planned for 2026. This signals that the music industry is moving toward licensing frameworks rather than purely litigation-based approaches.

Talent Contracts in the AI Era

Talent agreements are being fundamentally restructured to address AI. The core provisions now include digital replica rights (when and how a performer's likeness can be recreated digitally), voice cloning restrictions (whether and how a performer's voice can be synthesized), AI-generated performance terms (compensation and approval rights when AI replaces live performance), duration and scope limitations (how long digital replica rights last and what uses are permitted), and additional compensation triggers (when AI use of a performer's likeness requires separate payment). California's law effective January 1, 2025 prohibits contracts that create digital replicas unless specific exceptions apply — such as the person being represented by a lawyer or the digital replica not replacing work the person would otherwise perform. For entertainment lawyers, this means every talent agreement needs AI-specific provisions reviewed against both California law and emerging legislation in other states. The WGA and SAG-AFTRA agreements established AI guardrails for writers and performers in 2023, but individual talent contracts still need negotiation on the specifics — and the leverage dynamics vary by talent tier.

Content Licensing for AI Training

The UMG-Udio settlement represents the first major licensing model for AI training on entertainment content, and it will influence how other content categories handle AI use. The deal structure — license existing catalog, add guardrails for current use, develop a new jointly built service for 2026 — suggests the music industry sees AI as a revenue opportunity rather than purely a threat. But licensing frameworks for AI training face practical challenges. Who sets the rate? How do you audit compliance? What happens when an AI model trained on licensed music generates output that competes with the licensed catalog? These questions are being negotiated in real time, and the answers will vary by content type. News content (NYT vs. OpenAI), images (Getty vs. Stability), music (UMG-Udio), and video (the next frontier) will each develop different licensing structures reflecting the economics and copyright characteristics of each medium. For managing partners advising content companies, the strategic question is whether to license, litigate, or both — and the answer depends on the client's content library, competitive position, and revenue model.

AI-Generated Content and Copyright Registration

The Copyright Office's ongoing guidance on AI-generated works creates practical challenges for entertainment companies. Works generated entirely by AI without human creative authorship are not copyrightable. Works where a human makes creative choices using AI as a tool may be copyrightable — but only the human-authored elements receive protection. For entertainment companies producing content with AI assistance, this means careful documentation of human creative involvement at each stage of production. A film studio using AI for visual effects needs to demonstrate that human artists made creative decisions about the AI-generated elements. A music label releasing AI-assisted tracks needs to document the human producer's creative choices. A publisher using AI for content generation needs to identify which elements reflect human authorial judgment. The practical workflow: treat AI as a tool, document human creative choices at every stage, and register copyrights only for elements where human authorship is clearly demonstrated. This documentation requirement adds overhead, but it's essential for protecting the commercial value of AI-assisted creative works.

What Media and Entertainment Firms Should Do Now

Entertainment law practices need to build capabilities across four areas simultaneously. Copyright strategy: Advise content clients on whether to license their catalogs for AI training, litigate against unauthorized use, or develop hybrid approaches. The NYT, Getty, and UMG cases each represent different strategic choices. Talent representation: Rewrite talent agreements with AI-specific provisions covering digital replicas, voice cloning, and AI-generated performances. Track California's digital replica law and emerging legislation in other jurisdictions. Content licensing: Develop licensing frameworks for AI training that address rate-setting, compliance auditing, and competitive use restrictions. The UMG-Udio deal is a template, not the final answer. Copyright registration: Build workflows that document human creative involvement in AI-assisted content production, ensuring copyrightability of the resulting works. The firms that become known for AI-era entertainment expertise will capture the premium advisory work as every major entertainment company restructures its approach to AI.

The Bottom Line: The entertainment industry's AI legal framework is being written in real time through the NYT, Getty, and UMG cases. California's digital replica law is forcing talent agreement rewrites. The UMG-Udio settlement suggests licensing rather than pure litigation is the industry's direction. For managing partners, the opportunity is in building expertise across copyright strategy, talent representation, content licensing, and AI copyright registration — the firms that own this intersection will capture the premium advisory work for the next decade.

AI-Assisted Research. This piece was researched and written with AI assistance, reviewed and edited by Manu Ayala. For deeper takes and the perspective behind the research, follow me on LinkedIn or email me directly.