Universal Pictures warns Big Tech: don't train AI on our movies

Movie credits used to end with that familiar FBI warning telling people not to smuggle camcorders into theaters. Now, Universal Pictures is updating the fine print to warn off an entirely different kind of pirate: Silicon Valley.

Universal is rolling out a brand-new legal notice at the very end of its credits stating its films "may not be used to train AI." The clause first quietly appeared on How to Train Your Dragon in June, and has since been attached to Jurassic World Rebirth and Bad Guys 2.

The anti-scraping end credits

The text warns that the movie is protected under global laws and that unauthorized duplication could trigger civil liability and criminal prosecution. In several jurisdictions, it directly cites a 2019 European Union copyright directive that allows creators to explicitly reserve their rights and opt out of scientific text and data mining—a rule Big Tech spent years lobbying against.

A person familiar with the studio's thinking notes that slapping the language onto theatrical prints worldwide builds another line of defense against AI scraping.

Hollywood's paranoia is well-earned. By 2024, text-to-image models like Midjourney were already spitting out near-identical frames when fed everyday prompts. Ask for "Thanos Infinity War" and you get the purple villain looking ripped straight from Marvel promo material. Type in Top Gun: Maverick, and out comes Tom Cruise strapped into a cockpit. Midjourney could effortlessly duplicate styles from DreamWorks' Shrek, Pixar's Ratatouille, and Warner Bros.' The Lego Movie, giving Universal and Disney all the evidence they needed to sue the platform.

Targeting the copy step

The crux of these legal brawls usually centers on fair use, an area where copyright holders have struggled lately. Book authors recently lost fair use arguments in high-profile lawsuits against Meta and Amazon-backed Anthropic.

Universal's new strategy zeroes in on a different technical vulnerability: unauthorized copying. To feed frames into a training pipeline, AI companies have to create digital duplicates. If those copies were pulled from unverified or pirated sources, fair use gets much harder to claim. With statutory damages for willful infringement reaching up to $150,000 per title, the financial stakes are massive.

Meanwhile, the tech side is moving fast. On July 30, Edward Saatchi's Fable Studios announced an undisclosed investment from Amazon into Showrunner, an AI platform designed to let users prompt entire TV episodes into existence. Showrunner is already in talks with legacy studios, including Disney, aiming for an AI-generated theatrical release as early as 2026.

Why it matters

Universal's move is clever lawyering, even if it feels a bit old-school.

We think the real goal here isn't to scare scrapers off in real-time—web crawlers aren't sitting in theater seats reading the credits. Instead, it is about setting a legal trap. Tech companies love claiming ignorance, arguing that web scraping is fair game unless an owner clearly opts out. By embedding explicit opt-outs and citing EU statutes right on the master distribution prints, Universal strips away Big Tech's innocent-infringement defense. If Universal catches a model reproducing its frames, it can immediately push for maximum willful infringement damages of $150,000 per work.

At the same time, this highlights the weird identity crisis gripping Hollywood. Traditional studios are aggressively suing mid-tier AI generators like Midjourney to protect their existing IP, while simultaneously exploring licensing deals with startups like Showrunner to turn passive viewers into prompt-based creators.

Suing over training data might slow AI companies down in court, but it will not stop the shift toward generative video. In the end, the studios that win won't just be the ones with the best lawyers, but the ones that figure out how to license their universes before users simply generate their own.