Bringing Law and Order to the AI Wild West

To anyone who has been following AI copyright-related news lately, it is crystal clear that we are entering the next chapter of the AI and copyright story—the licensing phase.

The first few years of AI, which I’ll refer to as chapter one in the AI-copyright story, was exemplified by large AI companies refusing to license the copyrighted works they were training on. Instead, they claimed that licensing was impossible and that they needed free use of copyrighted works for national security reasons, to keep up with China, or any other preposterous rationales that were spun to play well in the media or with policymakers despite facts to the contrary. In essence, this chapter in the story was typified by a wild west approach, where large AI companies did whatever they pleased, defiant of any repercussions. 

But the creative community started to push back. There were lawsuits. Lots of them. (Close to seventy at this point.) There were also a few settlements. Eventually, we started seeing some license deals being made by certain AI companies. Those deals were forged mostly by smaller AI companies like ProRata.ai. and Bria, which understand that free market solutions and licensing supported by copyright law create a sustainable path forward for both the AI and creative sectors. Larger AI companies, on the other hand, tended to throw their weight around and not license much of anything, mistakenly treating the situation as a zero-sum game.

We’ve witnessed spikes in licensing, agreements, and other constructive partnerships and developments in response to different news, like when CloudFlare offered tools for news media publishers to make their news media content unavailable for AI scraping without permission and when news of the Anthropic settlement first broke. But, in many ways, those spikes pale in comparison to the explosion of licensing activities we have witnessed over the past several weeks.

If you missed the news of the AI copyright license deals made over just the past few weeks, here’s a quick rundown:

  • Disney/OpenAI Deal

Disney and OpenAI announced a three-year licensing agreement that will allow Sora, OpenAI’s video generation model, to generate short, user-prompted social videos that can be viewed and shared by fans, drawing on more than 200 Disney, Marvel, Pixar and Star Wars characters. As part of the agreement, some fan-inspired Sora short form videos will be available to stream on Disney+. As part of the agreement, Disney and OpenAI affirm a shared commitment to responsible use of AI that protects the safety of users and the rights of creators, including measures to guard against harmful or illegal output or violations of individuals’ rights in their voice and likeness. Also, as part of the agreement, Disney will make a $1 billion equity investment in OpenAI and receive warrants to purchase additional equity—a clear signal that both copyright owners and AI companies can thrive when they work together, and copyright laws are respected.

  • UMG/Udio Deal

Universal Music Group (UMG) announced that it settled its AI copyright infringement lawsuit against Udio. UMG was the first of the major music companies to do so and this news kickstarted the end-of-year domino effect of other deals in the music industry. In addition to the compensatory legal settlement, the ability to share music beyond the service was disabled, and the parties entered into training license agreements for UMG’s recorded music and publishing catalogues. According to the announcement, UMG and Udio will work on launching a new subscription service in 2026 for a generative AI service fully re-trained on only licensed music. The license is structured as artists’ opt-in.

  • Major Record Labels/Klay

The three global music companies—Universal Music Group (UMG), Sony Music Entertainment (SME), and Warner Music Group (WMG)—announced they had each licensed music to the responsible generative AI music service Klay. The startup’s large music model will only be trained on licensed music, and the companies are working to develop a “pioneering commercial, ethical foundational model.”

  • WMG/Udio Deal

Warner Music Group (WMG) announced that it settled its AI copyright infringement lawsuit against Udio. The settlement agreement includes a licensing deal where WMG artists can opt-in to license their works to develop Udio’s new subscription service, which will be launched next year.

  • WMG/Suno Deal

Warner Music Group (WMG) and Suno announced that they reached an agreement to settle the lawsuit filed by WMG and other record labels against Suno over the unlicensed use of music from the record labels repertoire to train Suno’s AI model. According to the announcement, Suno will launch an entirely new models in 2026 that are “more advanced and licensed” while current models will be phased out. Free tier users of the new AI model will not be allowed to download audio outputs but instead can play or share them. The announcement also makes clear that “[a]rtists and songwriters will have full control over whether and how their names, images, likenesses, voices, and compositions are used in new AI-generated music.”

  • News Media Publishers/Meta Deal

Meta announced that it has struck multi-year AI partnership deals with a variety of news media organizations including CNN, Fox News, Fox Sports, Le Monde Group, the People Inc., The Daily Caller, The Washington Examiner, and the USA TODAY Network, which would offer these publishers’ news media content on Meta AI, the company’s AI chatbot. Users can prompt Meta AI to generate responses which will include links to articles, ensuring that users know the legitimate origin of the content and can visit the outlets’ websites.

  • UMG and WMG/Stability AI Deal

Universal Music Group (UMG) announced that it entered into a partnership with Stability AI to develop “professional-grade tools that enable artists, songwriters, and producers to experiment, compose, and produce using ethically trained models.” Warner Music Group (WMG) announced its own license with Stability AI soon thereafter.

  • Major Music Publishers/Musixmatch Deal

Musixmatch announced that it has signed AI licensing deals with Sony Music Publishing, Universal Music Publishing Group, and Warner Chappell Music, to use the publishers’ musical works to develop AI tools. The licensing deals were reportedly done on a trial basis and may include the development of a large language model that would analyze song lyrics and metadata to deliver answers for specific user questions and prompts.

  • Getty Images/Perplexity Deal

Getty Images and Perplexity announced a global multi-year licensing agreement covering the display of images from Getty Images across Perplexity’s AI services and tools. The agreement allows Perplexity to provide high quality creative and editorial imagery, including image credits with links to sources.

Wow. This is not even close to a comprehensive list of AI copyright licenses, but it provides some sense of the number and significance of the AI deals made by copyright owners over the past several weeks. (A more comprehensive list of the deals between AI companies and copyright owners (including a first early wave of deals, many of which are with news publishers) can be found here.)

These deals are a huge win. On one side of the AI-copyright disputes are AI companies who have argued that they should be able to train on copyright owners’ works without permission. On the other side are copyright owners who have argued that AI companies must license their works when used for training purposes. In each deal listed above (and the many other deals not listed), the end result is that the AI company is licensing the works for use by the AI company. The endgame here, has always been a business model based on licensing. Whatever the AI companies said about licensing being unachievable has been proven wrong by these deals. A free market aims to bring buyers and sellers, rightsholders and users together—and thanks to the copyright framework that supports free market licensing, we are now starting to see that goal being achieved.

There can be no doubt that, standing alone, these deals not only benefit the copyright owners involved in the deals, but also represents an enormous win for the entire creative community more generally. These deals unambiguously refute the big AI companies’ mantra that AI licensing is not possible and should result in even more big AI companies coming to the table to negotiate additional deals in the future as they compete for high-quality copyrighted works for various AI use and applications.

These deals are also win for the AI companies involved. Not only do these deals erase a major risk of liability for them, but importantly, they also provide frameworks to ensure that their model output is safe and less subject to hallucination. In addition, these deals also give the AI companies involved a competitive advantage over their competition that does not have such a deal in place.

Perhaps, most importantly, these AI copyright licensing activities show that there is a cognizable licensing market under a fourth fair use factor analysis and that market is harmed when AI companies steal, rather than license, copyrighted works. These deals now serve as direct evidence of the existence of those markets, and the future of these markets, in the close to seventy AI copyright infringement cases that have been filed in federal courts throughout the country (many of which are class action suits involving individual creators). So, while companies like UMG, Disney, and USA Today directly benefit from these deals, so too will all the individual creators who are part of the class action lawsuits that have been brought. Judges who in the past may have found licensing markets difficult to recognize and reconcile as part of a fourth factor analysis will find it impossible to deny their ubiquity and significance.

What is most significant about these AI copyright licensing deals isn’t just the volume of deals being struck, but the components of the deals. Most of these deals include four important elements:

  • Compensation: Although announcements related to these deals typically are silent on the amount of money being paid by the AI company to the copyright owner, there is clearly a substantial amount being paid to the compensate the copyright owner for use of their works for training.
  • Consent: These AI copyright license agreements ensure that models are trained on works based on permissions from the copyright owner. It has long been the goal of the copyright community to ensure that AI companies seek and obtain consent from copyright owners when using their works for training. And that is exactly what these agreements require.
  • Control: In deals where the training on works of individual creators is involved, these deals guarantee creators the ability to opt-in before their works can be ingested for training. This means that creators get to decide whether and how their works are used. In relevant agreements, these creators also get to decide whether their name, image, and voice are used. While in other deals, there is a clear statement that name, image, and voice may not be used at all. Significantly, these deals also delineate how the copyrighted works at issue can be used and for what purpose.
  • Collaboration: Many of these agreements, such as the OpenAI, Suno, and Udio agreements are more than just license agreements, they are collaborative partnerships between AI companies and copyright owners in which the AI company agrees to train on and/or generate only licensed, authorized content, and the two entities establish a continuous business relationship going forward in which both stand to benefit by working together on developing technology. It’s unequivocal evidence that the future of generative AI can be win-win for copyright owners and AI companies.

These deals are far from perfect, and there are a lot of details that we simply do not know about yet or have not been ironed out between the parties yet. And because of this, in conjunction with a history that certain creators or others may have with certain companies or industries, there may be some fear, apprehension, or anger by certain creators with these deals.

For example, a couple of comments on social media questioned whether AI companies should be rewarded with a major deal when, for the past several years, they have been copying billions of copyrighted works without payment or permission. Others have rightfully questioned why deals haven’t been made yet with individual creators and whether individual creators will get their fair share from the industry deals.

To those people, I urge patience and an understanding that we are at the very start of this second phase in the AI-copyright timeline. In time, those deals will come. And at the rate these deals are being announced, it may not be long either. As more deals are reached between AI companies and copyright owners, we begin to establish a strong foundation built, on compensation, consent, control, and sometimes collaboration.

What we are witnessing right now is a reining in of the wild west. These deals demonstrate a steady shift away from the approach where AI companies use whatever creativities they want and towards a world in which AI companies license the works they are trained based on our copyright law framework. That is a big win for copyright owners, for AI companies, and the United States.


If you aren’t already a member of the Copyright Alliance, you can join today by completing our Individual Creator Members membership form! Members gain access to monthly newsletters, educational webinars, and so much more — all for free!

get blog updates