Bad News for Movie Studios: Authors Just Lost on a Key Issue In a Major AI Lawsuit

by

Transformative.

In simple terms, this federal court has labeled Amazon-supported Anthropic’s method of learning its AI system from millions of books found online as such. This is the initial decision addressing this topic, and it will likely set a precedent for future cases involving similar situations. The verdict may cause concern among studios, as some have become involved in the dispute over this groundbreaking technology.

As an enthusiast, I’m intrigued by the central question these cases revolve around: Does the principle of fair use in intellectual property law apply to AI companies, enabling them to utilize copyrighted works without a license for their development? In this particular case, it seems that Anthropic has a strong legal standing, at least when it comes to the training aspect.

U.S. District Judge William Alsup stated that the technology is “one of the most groundbreaking we’re likely to witness during our lifetime.

Despite purchasing the books it had previously illegally downloaded, Anthropic will stand trial for downloading seven million books unlawfully to create a library used for training purposes. The court found that this action does not exonerate Anthropic from liability. The company could be liable for hundreds of millions of dollars in damages due to the decision, which may also result in similar payouts to Disney and Universal if they can prove during discovery how Midjourney obtained copies of thousands of films that were later used to train its image generator.

Previously, writers initiated a legal action against Anthropic, alleging that the organization had unlawfully obtained and duplicated their books for use in training its AI model, Claude. Rather than attempting to have the case thrown out, Anthropic opted to bypass this step and directly address the question of fair use instead.

The court’s decision states that authors cannot prevent Anthropic from utilizing their works for technological training, similar to how no one can be stopped from reading a book to educate themselves on writing skills.

The instruction states that everyone not only reads existing texts but also creates new ones. However, acquiring a text might require payment initially. The idea of requiring payment every single time someone reads the book, recalls it from memory, or uses it as a reference when composing something new is inconceivable.

In simpler terms, the court found that if someone reads and learns from modern-day classics, imitating a mix of their writing styles, it does not count as copyright infringement. Just like any aspiring writer, Anthropic’s technology uses these works as inspiration, not to replicate or replace them, but to generate something completely new, the order states.

The outcomes Disney and Universal didn’t anticipate or desire in their lawsuit against Midjourney for copyright infringement is causing them concern. They fear that Judge Alsup’s assessment might influence the judge handling their case, potentially undermining the development of a groundbreaking technology. On a broader scale, there’s a possibility that AI video generators, such as Sora, may be seen as condensing all movies into entirely new creations.

According to Daniel Barsky, a legal expert in intellectual property at Holland & Knight, it’s expected that everyone developing AI models will reference this Anthropic decision as evidence when they argue that the principle of fair use is applicable when working with huge datasets to build fundamental models.

It’s worth mentioning that the authors didn’t claim Anthropic’s responses copied their work. In fact, had they made such a claim, they would have likely lost due to the court’s decision that safeguards prevent any infringing content from reaching users. This was likened to Google setting restrictions on how much text from one book a user can view via its Google Books service, thus preventing the search function from being misused as a means to access entire books for free.

In the scenario where user-seen outputs were violating copyrights, authors would have a distinct legal issue. However, if those outputs were to breach copyrights in the future, authors could initiate such a lawsuit. But that’s not what we’re dealing with here.

However, it’s possible that Midjourney generates images that are almost identical to certain film frames in specific scenarios.

As a gamer, when I asked Midjourney for an image related to “Thanos Infinity War”, it delivered a highly realistic graphic straight out of the Marvel movie or promotional materials. Similarly, when I requested a shot from “Top Gun: Maverick”, it provided an image of Tom Cruise in a fighter jet cockpit just like in the film. This AI tool can seemingly mimic various animation styles, creating stunningly accurate characters from a vast array of titles, such as DreamWorks’ “Shrek”, Pixar’s “Ratatouille”, and Warner Bros.’ “The Lego Movie”.

The complaint alleges that since Midjourney creates replicas and adaptations of movies from Disney and Universal, it is clear that the company, without the necessary approval or awareness, has been using and adapting plaintiffs’ copyrighted materials to refine and enhance its technology.

Additionally under consideration: The speculation that Midjourney unlawfully obtained the studios’ films. In his June 23 decision, Judge Alsup determined that Anthropic’s act of illegally downloading seven million books for the purpose of creating a training library does not fall under fair use. He stated that the company could have instead purchased the copies. This form of piracy, the court declared, is “fundamentally and irremediably infringing.

If intentional copyright violations lead to statutory damages as high as $150,000 per piece of work, it’s possible that substantial financial penalties could be imposed.

Read More

2025-06-25 03:24