The AI Copyright Wars

Background

The question of whether AI models trained on copyrighted works constitute infringement was theoretical until it wasn't. Generative AI crossed the line from research curiosity to commercial product in 2022, and the lawsuits followed almost immediately.

The core tension is structural. Large language models and image diffusion models require vast training datasets. Those datasets inevitably contain copyrighted material: books, images, music, journalism. The companies building these models argue that training constitutes fair use, a transformative process that produces new works rather than copying existing ones. Rights holders counter that the outputs directly compete with their original works, generated from a foundation of unpaid labor.

What makes the AI copyright question different from previous technology disputes like Napster, YouTube, and Google Books is scale. A single model can absorb millions of works, and its outputs can substitute for any of them. The economic displacement is not theoretical.

Aftermath

The legal campaign has escalated across every creative medium.

Visual art was the first battleground. In January 2023, artists Sarah Andersen, Kelly McKernan, and Karla Ortiz filed a class-action lawsuit against Stability AI, Midjourney, and DeviantArt. Three weeks later, Getty Images sued Stability AI in both US and UK courts, alleging infringement of over 12 million photographs.

Journalism followed at the end of 2023. The New York Times filed what became the highest-profile case: a federal lawsuit against OpenAI and Microsoft alleging millions of articles were used without permission to train ChatGPT and Bing Chat. In October 2024, Dow Jones and the New York Post sued Perplexity for reproducing paywalled content with minimal attribution. By December 2025, the Times had also sued Perplexity.

Music entered the arena in June 2024, when the RIAA filed simultaneous lawsuits against Suno and Udio, the two leading AI music generators, alleging mass copyright infringement in training data. In October 2025, Universal Music Group settled with Udio, agreeing to jointly develop a licensed AI music creation service. It was the first major settlement in the AI copyright wars.

Film and video became the latest front in February 2026, when ByteDance's Seedance 2.0 video generator went viral, producing clips featuring Disney characters, Marvel heroes, and anime icons like Ultraman and Detective Conan without authorization. Disney issued a cease-and-desist letter accusing ByteDance of a "virtual smash-and-grab." SAG-AFTRA and the MPA condemned the tool. Japan's Cabinet Office opened a formal copyright investigation. ByteDance pledged to strengthen safeguards.

Industry Impact

As of February 2026, no US court has issued a definitive ruling on whether AI training constitutes fair use. The cases are proceeding through discovery and motions. But the legal uncertainty has already reshaped the industry.

The licensing pivot. Some companies have begun licensing content proactively. OpenAI signed deals with the Associated Press and Axel Springer. Perplexity launched a Publishers Program to share revenue. These deals implicitly acknowledge that training on copyrighted material carries legal risk.

The geographic divergence. Japan's copyright law initially appeared more permissive toward AI training, but the Seedance investigation suggests enforcement is tightening. The EU's AI Act includes provisions for training data transparency. The US remains the most uncertain jurisdiction.

The music precedent. The UMG-Udio settlement is the first model for how AI companies and rights holders might coexist: joint development with licensing. Whether this extends to text and images remains to be seen.

The opt-out illusion. robots.txt and opt-out mechanisms have proven inadequate. Multiple lawsuits allege that AI companies ignored these signals. The Perplexity cases specifically cite evidence of scraping past hard-blocks.

The displacement question. Courts will ultimately need to address not just whether training is fair use, but whether AI outputs that compete with original works constitute market substitution. That is the fourth factor in fair use analysis, and historically the most decisive.