Why generative AI litigation in Hollywood is ripe for mediated settlement

What do Scarlett Johansson, Hollywood studios, Tony Robbins, major record labels, Sarah Silverman and a spectrum of other artists and media companies have in common? They're all players in the wave of legal claims against new AI companies.

From the first silent films up through today's world of streaming, Hollywood's creative visionaries and business innovators have embraced new technologies while also defending their existing rights. This evolution culminates now of course with Generative Artificial Intelligence tools (GAI). Unsurprisingly, players in entertainment are once again at the forefront of change and policing abuses.

Challenging existing paradigms comes with a price, of course. Legal issues arise as new tools emerge, push boundaries and disrupt dominant models. As an incubator for groundbreaking lawsuits--driven by celebrities, famous intellectual property (IP), and the publicity they attract--Hollywood litigation is not only juicy but also makes new law.

Most media cases eventually settle though, and for good reason. They are complex and costly, involve IP concepts that challenge judges and juries, and their novel issues risk unpredictable outcomes. Adverse judgments also threaten reputational harm while setting bad precedent. And pragmatically, in this relationship business, parties in conflict today will often be working together tomorrow.

For all the same reasons, disputes in the AI wars are starting to settle, often with the aid of mediators. This article explores some of the most notable conflicts in this new battleground and what we know about how they resolved. Why does it matter? Their emerging patterns and precedent-setting licensing frameworks provide a roadmap for future negotiations as this landscape evolves.

The $1.5B settlement that shook the world, and the continuing copyright wars

The most staggering GAI settlement came in August 2025 in Bartz v. Anthropic, where a class of over 300,000 authors sued Anthropic for copyright infringement stemming from its platform called "Claude." The authors claimed Anthropic infringed their copyrights in over 500,000 titles by (i) copying their books to train its large language model and (ii) allowing users to generate substantially similar infringing outputs (with comparable theories asserted in pending cases by The New York Times and others).  

In a landmark opinion from the Northern District of California, Senior U.S. District Judge William Alsup viewed GAI's implications for creativity somewhat optimistically, found that fair use barred some but not all claims (where the copied books were also pirated), and certified the class. The implication? Anthropic faced, by some estimates, a company-killing statutory damages award at trial in the high hundreds of millions to possibly trillions for willful infringement. The consequence? With the help of a mediator (Layn Phillips), the parties reached a $1.5 billion settlement touted as the largest in copyright history. Notably, the settlement only covers past infringement, leaving the authors to shape an emerging licensing market.

Uncertainty, however, is still the name of the game. Why? In a twist only Hollywood could write, just weeks later, a case brought by Sarah Silverman and other artists--involving comparable facts--met the opposite fate before a different judge in the same courthouse. That case, Khadrey v. Meta, involves Meta's GAI model called "Llama" and similar allegations that it trained on datasets of copied books and created infringing outputs. U.S. District Judge Vince Chhabria, by contrast, viewed GAI's implications for creativity more skeptically. Yet while he nonetheless also reached essentially similar conclusions on fair use as Judge Alsup (but for different reasons), he dismissed the key claims based on the record before him in an unprecedented analysis (say some legal experts).

This summary presents a simplified distillation of highly complex rulings, of course. For our purposes, however, the point is that seemingly analogous cases can turn out very differently depending on the courtroom. The moral? These GAI cases are prime candidates for negotiated resolution due to their complex issues, the extraordinary cost to litigate their technical claims, and their unpredictable outcomes.

To underscore this proposition, a mediation is now teed up in the landmark lawsuit brought in June 2025 by Disney, NBCUniversal and Warner Bros against the GAI platform Midjourney. The first such suit brought by Hollywood studios alleges Midjourney ripped off their famous IP to train its model and generates outputs that copy their famous characters (from Buzz Lightyear to Darth Vader and many more). Midjourney, of course, offers many nuanced and complex defenses, and with a mediation completion date in August 2026, we'll soon see if settlement is in the cards and what impact, if any, Disney and OpenAI's precedent-setting licensing partnership announced in December 2025 may have on its contours.

The court of public opinion--As judge, jury and executioner

It is said life imitates art in Hollywood, and it does in Silicon Valley now too. Exhibit 1? The most public AI war that wasn't--Scarlett Johansson's tussle with OpenAI over her voice.

In Spike Jonze's 2013 film "Her," ScarJo famously voices a faceless AI agent opposite Joaquin Phoenix. Fast forward to 2024, when OpenAI's founder Sam Altman, who has said "Her" is his favorite film, courted Johansson to be the voice of ChatGPT. Though she declined, on the eve of the bot's launch, she discovered OpenAI planned to use an artificially generated voice that was uncannily similar to hers.

Shocked, angered and in disbelief (her words), Johansson sent cease & desist letters alleging her likeness had been misappropriated and demanding OpenAI detail its process in developing the voice. Quickly changing course, while avoiding litigation and possibly sensitive discovery, OpenAI announced it was shelving the voice while exploring "questions about how we choose the voices in ChatGPT" (without further comment).

Sound familiar? You might recall OpenAI's other famous course correction after launching "Sora2" with a default policy that allowed users to generate copyrighted characters unless the rightsholder explicitly "opted out." By contrast, likenesses of real people could not be generated unless they had "opted in." The intense backlash lasted 72 hours before OpenAI shifted to an "opt in" for copyrighted content too. In the meantime, Sora2 had quickly climbed to No. 1 on the App Store.

The takeaway? Celebrities and rightsholders will not tolerate exploitation of their likenesses and intellectual property via GAI without consent and compensation. Equally clear? Some AI companies will push to remain competitive until made to change course.

The new frontier in misappropriation and trademark theories

AI cases that actually made it to litigation on similar grounds are also now starting to settle. Among them is Vacker v. Eleven Labs, widely heralded as the first major AI case to settle, which did so on confidential terms with the help of a mediator (Jeff Kichaven).

In August 2024, two accomplished voice-over artists from Texas and New York sued the start-up Eleven Labs, whose GAI tool creates audio content and voices used for narration. The artists, one male and one female, claimed under their state laws that Eleven Labs copied audiobooks they had narrated and then misappropriated their voices/likenesses to create very similar-sounding artificial voices called "Adam" and "Bella." The authors of the copied books also sued under the Digital Millenium Copyright Act (DMCA), claiming their digitally embedded Copyright Management Information (CMI) had been stripped from the works to copy them.

The docket reveals why early settlement can make sense in AI cases. The heated litigation included (i) a motion to dismiss based on technical substantive defenses (including that plaintiffs admittedly were not celebrities) coupled with standing and jurisdictional challenges; (ii) a motion to transfer venue; and (iii) a claim by the defendants for their attorneys' fees under New York's SLAPP statute. A year after filing, and facing significant exposure, costs and risks on both sides, a mediated settlement in August 2025 offered the parties finality and self-determination instead of expensive uncertainty.

Around the same time, in July 2025, the indomitable Tony Robbins sued InnoLeap AI and Mira Muse, the companies behind the "YesChat.AI" platform, for their unauthorized creation of bots with names like "Talk to Tony Robbins," "Tony Robbins GPT" and "Tony Robbins Español GPT." Robbins alleged they ingested his copyrighted content, digitally recreated his likeness and masqueraded as his persona to dispense his style of advice to paid subscribers. In a sign of the times, he claimed damage from the bots competing with his own AI clone that he already offered on his own site.

The suit offers a case study in a legal strategy that paid off. Despite the alleged heist of his copyrighted works, Robbins only alleged the more established celebrity-based trademark and misappropriation claims, thereby sidestepping the AI copyright wars. The companies defaulted, then surfaced, then quickly entered into a stipulated judgment--agreeing to cease and desist, destroy all copies of the unauthorized content and pay $1M in damages.

This public settlement, driven presumably by the costs of fighting a deep-pocketed Robbins under California's celebrity friendly laws, sends a clear message to those who might have similar designs on a celebrity's famous likeness.

Music library litigation and the emerging fault lines

Some of the music industry's major record labels have also reached ground-breaking GAI settlements in their still contested landscape, while some have not. In June 2024, Sony Music, UMG and Warner Records sued Udio and Suno--GAI companies whose platforms allow users to create audio content. The copyright claims alleged mass unauthorized copying of recordings to train the music-generating models, and that the outputs closely mimic the sound and style of famous artists and compete with their music.

In late October 2025, after over a year of litigation and in what some consider a strategic pivot, Udio and UMG settled and announced a licensing partnership. The "walled garden" approach lets users generate content based on licensed music under usage and distribution constraints. They also announced payments for artists tied to training and outputs, but those details and their consent rights are not yet public.

A few weeks later, Warner Music also settled with Udio and announced a similar licensing arrangement. A week after that, Warner Music and Suno settled and announced another licensing arrangement (involving the bigger alleged offender this time). Meanwhile, UMG fights on against Suno, and Sony steadfastly maintains its actions against both platforms.

Not surprisingly, artists and their representatives have expressed support for deals that provide them with control and compensation, but the devil is in the unknown details, and they are still reserving judgment.

In closing

To settle or not to settle? That is always the question. The calculus depends on what is at stake, what it costs to vindicate your rights and what risks you face. GAI cases involve critical IP at the core of these litigants' ability to survive in the market. Their high stakes will therefore only spawn more disputes where cost effective, principled resolution will continue to make sense.

As more settlements take shape, and as the licensing landscape evolves, new options are emerging for creativity in the negotiations. Included among them are novel compensation models; protections for creators and their publicity rights; cross-licensing frameworks for technology and IP; guardrails to ensure quality control; and even opportunities to invest in AI partners and exploit user-generated content.

Stay tuned as the AI wars, and now the peace accords, continue to make history.