Sync Summit Weekly Update: July 28, 2026

What’s in this week’s letter?

Hello and welcome to the new newsletter format

A word from your sponsor – Upcoming Sync Summit events, classes and more.

This Week’s Deep Think: The Suno Hack, Fair Use, Lawsuits and Building a Fair Economy for Music in the AI Era

First and foremost, I want to wish you a great start to your week. I hope you’re doing great personally and professionally, and I hope that what we do at Sync Summit brings real value to you and your work. 

Today, I’d like to announce a change to the format of the emails we send out to you and our social media posts.  I’ve had a good look at the quality of our emails and social media outreach and I believe we can and should do more than use our platform primarily as a way to market our events, courses, listening sessions and consulting services.  

With all the changes and uncertainty in our industry, its crucial we use our platform to provide analysis and understanding of the impact and evolution of technology and its effects – good and bad – on the music business in general and the music for media sector specifically.The new format reflects this. 

Along with the usual marketing of Sync Summit’s services, we’re going to do a weekly Deep Think article that tackles one issue facing the industry.  

This week’s article focuses on the implications of the recent Suno hack, the company’s use of copyrighted music works for training its LLM, how this impacts Suno’s fair use defense in its ongoing litigation, and how we as an industry can build a way to ensure artists and rights holders are paid for the usage of their works to train AI systems. 

In addition, we’re going to start holding weekly online conversations with leaders in AI, tech and sync to discuss industry developments, changes, challenges and new tools you can use for your work in music for media. We’ll announce these in the near future.In the meantime, thank you for everything you do to make our industry and our world a better place. 

Best,

Mark Frieser

CEO, Sync Summit 

A word from your sponsor: Upcoming Sync Summit events, courses and listening sessions. 

Events: We have three events coming up:  

First, we’re announcing the LA edition of our AI and Sync Symposium, which is taking place at our home in Santa Monica, The Recording Club on September 16.  This is a one day event bringing you together with the companies, industry decision makers, technologists, music supervisors, studios, agencies, brand and networks creating the future of the music in media business.  Tickets are 50% off until July 31 – click here to learn more and sign up.

Second, there’s the NY AdSync Summit on October 1-2 and the LA Sync Summit on February 8-9. These are our flagship industry events gathering music supervisors, sync agents, artists and industry executives to network and do business. Click here to learn more about NY and here to learn more about LA – sign up to attend at 50% off until July 31.

Classes:  We have two classes coming up – and this is the last time they will be available:

The Sync Summit Camp, starting August 2: This is an intensive, personalized six-week course combining listening sessions and practical information on how to develop your sync practice successfully as an artist, a sync agent or a music supervisor. The Cost is $599 – click here to learn more about all the benefits and sign up.

The Music, Brands and Ads Course starting September 17:  This is an comprehensive course taught by music Branding expert Joshua Rabinowitz and Mark Frieser that provides you with deep insight and practical information on how to successfully work with ad agencies and brands.  The cost is $999 – click here to learn more and sign up. 

Annual membership Program.  Our annual membership program, which includes access to all our courses, listening sessions, events and monthly consulting time is open for enrolment until July 31 at the cost of $2000 – a huge savings on the cost of all of these benefits if purchased separately. After the 31st, we’re closing enrolment forever and the program will no longer be available.  Click here to learn more and sign up for membership while you still can. 

Supervisor Listening Sessions:  We have four sessions coming up this month where you can meet and present your music to leading music supervisors.  This month’s guests include Music Supervisors Julian Drucker, Joshua Rabinowitz, Yael Meyer and The What’s Up Pitches Production Team and Sync Agency. Click here for more info and to sign up.

And now… 

The Deep Think: The Suno Hack, Fair Use, Lawsuits and Building a Fair Economy for Music in the AI Era

So, what’s this about Suno getting hacked?

Earlier this month, the hacker got hacked when Suno, the generative AI music company’s source code was laid bare and the results reveal massive unauthorised scraping of decades worth of musical data from YouTube, Deezer, Genius, the catalogues of production music libraries Jamendo, Pond5 (2.5 million tracks), Freesound as well as multiple RSS feeds. This is in addition to the millions of songs the company trained its LLM on:“essentially all music files of reasonable quality that are accessible on the open internet”

According to to the breakdown provided in a 404 Media article, Suno ingested during one period in 2024 over 2 million YouTube clips to train its model, and according to a file about different datasets Suno created, there were “113,879 hours of youtube_music,” “17,615 hours of genius_hq,” “410 hours of free sound,” “19,514 hours of imslp,” “3,726 hours of jamendo,” “62,117 hours of pond5_music,” “12,287 hours of deezer,” “152,162 hours of ytm_tagged,” and “103 hours of musescore_lyrics.” 
This is a massive amount of training data derived from, unauthorised sources, and represents the tip of the iceberg of the training data aggregate in Suno’s dataset which includes podcasts, songs from major labels, publishers and independent artists – all without the permission of any of the rights holders.

Suno’s justification for using this material?  Fair use.

The fact that Suno is utilising massive amounts of copyrighted material without authorisation to train its LLM is no surprise, it’s been a known known for years.  

This is why they’ve been sued by Sony/Universal Music Group and German Performance Rights Organization (PRO) GEMA, and will likely be sued by other PROs, publishers and labels in the near future. 

But, they’re not just using copyrighted material without authorisation, they’re using the results of the training so their LLM can then create music based on prompts provided by users.  

In other words, they’re taking your music to train their tool to make music that competes directly with your music for streams, for syncs, for audience and ultimately for sales and subscribers. 

UGH!

And what’s Suno’s defense? Simply put, that they’re protected under the Fair Use doctrine. Here’s their reasoning for review from a 2024 court filing:

“It is no secret that the tens of millions of recordings that Suno’s model was trained on presumably included recordings whose rights are owned by the Plaintiffs in this case,” Suno was trained on “as many [recordings] as can be located … Accordingly, Suno’s training data includes essentially all music files of reasonable quality that are accessible on the open internet, abiding by paywalls, password protections, and the like, combined with similarly available text descriptions.” 

So what they are saying is that because something is available on the open Internet, it’s usable under the fair use doctrine for training purposes because it is a transformative use that brings new meaning to the original works, and that the usage of this material will not have an impact on the market for the original music.

This argument fails completely for a number of reasons. 

As someone who’s worked in IP for decades, I’m very  familiar with what comprises fair use and what the aspects of fair use are, so let’s go through all the aspects of fair use and see if Suno usage case falls under fair use:

What is the purpose of the use? 

 Is it commercial or non-commercial? It is obviously commercial, so their claim fails on that aspect.  Fail

 Is it Transformative?  This is a key tenet of their supposition. Does the usage add new meaning of purpose to the underlying material? Suno’s claim is that their usage is transformative – it is being used to create a new meaning or purpose from the material.  But is it? Suno is using the material to train their LLM to create new material (music) that is similar to those derived from the underlying copyrighted material, so this claim fails both on purpose and meaning. Weak argument and a Fail

What is the nature of the work?  Creative or factual?

As the training material is creative works, there’s a very weak to nonexistent case for fair usage.  Think of it this way.  If I quote in this article, as I did, 404.com‘s detailing of the hacked Suno comment’s file statistical information, and give them attribution, that’s fair use. 

On the other hand, if I copy and pasted an article from their website, that would not be considered fair use as an article in its entirety is a creative work. And the latter is what Suno did, writ large – utilising creative work on a massive scale for repurposing is not fair use.  So their claim fails on this aspect. Fail

Amount of material and the qualitative aspect of the material used?

Quantity: Obviously, Suno fails in this aspect because they are utilising untold millions of pieces of musical content to train their AI. Fail

Qualitative Aspect: This means whether the training data includes elements that are key components of the original datasets like actual vocals, lyrics, instrumentation, melody, rhythm rather than associated elements like descriptive metadata and ownership information.  If the answer is yes, which it is, then the argument this is fair use is falls apart. Fail

Effect on the Market.

What is the impact on the current market? Is Suno’s usage of copyrighted material harming the current or potential market for the current works? If so, the argument for fair use is weak to non-existent. Obviously, Suno’s music creation tool is materially and negatively affecting the current and future market for the copyrighted materials it’s using to train its AI, so Suno fails on this aspect of fair use as well. Fail

Application of Fair Use.  Normally, fair use is invoked for usage of copyrighted materials in one of three cases: 

Criticism and Commentary: Quoting a work to comment on or critique it.

Teaching and Scholarship: Excerpts for educational purposes.

News Reporting: Summarisation or quoting of works in news articles.

As none of these three cases of the application of fair use apply to Suno’s usage of copyrighted data to train their AI, their argument fails on this aspect as well. Fail. 

Final Analysis: Why does Suno’s fair use argument fail?

Based on the standard, acknowledged applications of fair use, Suno’s argument for the usage of copyrighted material to train their LLM falls outside every criteria of fair use and represents a massive, unauthorised usage of copyrighted material, and the data from the recent hack only reinforces this conclusion. 

Creating and building a business based on the unauthorised use of copyrighted materials to train an AI that is a purpose-built enterprise for profit that competes with and to a degree supplants the market of the current rights holders and creators is the exact opposite of fair use.  At the end of the day, their fair use argument will fail, and they will likely have to share revenue or equity with rights holders as part of any final judgement to avoid being shut down. 

The Suno hack data dump only reinforces the argument that their massive scraping of copyrighted data is the opposite of fair use.  And, with the GEMA (Germany’s performance rights organization) vs. Suno ruling coming this week, the hack couldn’t have come for a worse time for the company. 

I say this because Suno’s defense in the GEMA case, where they are accused of unauthorised reproduction, storage and commercial exploitation of sound recordings to train their AI model (link: https://www.gema.de/en/w/press-release-lawsuit-against-suno), is that because their system analyses mathematical and musical patterns rather than storing actual files, and that output is shaped by user prompts, Suno’s actions fall under fair use and text/data-mining.

Is this a worthy defense?  I don’t think so for two reasons. 

First, I know from personal experience you can prompt Suno to create songs that sound like specific artists, which belies that is is not just analysing mathematical and musical patterns, but has knowledge of specific artists and their sound built into its system. Second, for all the reasons I stated above, their data-mining doesn’t fall under fair usage in aspect or application. 

We’ll see on Friday the 31st how things shake out in the GEMA vs. Suno case, but this, along with the eventual resolution of the Sony/UMG case are, in my estimation, only the first of many litigious challenges Suno will face.  

Specifically, we’ve yet to hear from the other PROs around the world, let alone the major publishers, who are no doubt soon to litigate against the usage of their catalogues of compositions and lyrics for training material, along with the possibility of companies that were part of the data dump like Genius, Jamendo, Pond5 and others. In other words, the current litigation is only the tip of the iceberg of challenges Suno is facing in its usage of copyrighted materials for training.

And, this is all part of a much larger conversation around how data is being used to train AI in general and the types of litigation we’re going to see around generative AI music services in particular. Specifically:

Training Rights: Can an AI company copy copyrighted works to train a model without permission? This is what the Suno and Udio lawsuits are primarily about.

Output Rights: When an AI generates music that resembles an existing work, who is liable? This area is at the heart of the GEMA vs. Suno lawsuit, and will likely generate a second wave of further litigation.

Transparency and provenance: Can rights holders know whether their works were used? Can AI companies prove where training data came from? This is something that several operational AI services, like sureel.ai are working on right now, and it is crucial, because by knowing if your works were used to train an AI, how much and how often, it creates a possible path to ongoing artist and rights holder remuneration.

It’s transparency and provenance – knowing what material trained the AI and by how much that’s crucial to the long-term viability and remuneration of content creators and owners in the entrainment industry.  

In the short term, figuring out how and under what terms Suno and others can use copyrighted material and what liability exists, and to who, when a generative AI’s work mimics an existing work will be settled with a combination of money, equity and guardrails. But only by establishing a way to determine how much of who’s copyrighted material was used in the training and the output of generative AI music systems, can we create a reliable path to remuneration for content creators and owners. 

And this is where operational AI systems can help. An operational AI system that has been trained on… just hear me out… massive amounts of copyrighted data will be able to infer and derive the amount and types of music that has been been used to create a piece of AI generated music. From there, the output detailing what was used and by how much can be used as a basis for a future remuneration to rights holders.  

However, the current situation dictates either the generational AI companies would have to opt-in voluntarily to such a system or have it imposed on them legally.  My prediction is that a combination of enforcement and voluntary opt-in will happen within the next two-to-five years.  

For generative AI to be used in any salient way in audiovisual media, this has to happen.  

Sure, there are some ethical generative AI systems being used in media now, mostly in ads and social media, but for studios, networks, broadcasters, agencies and brands to use AI in their productions on a larger scale, a system needs to be established the mitigates risk of litigation and insures a method of remuneration to the underlying rights holders of training material.

Specifically these companies need to know whether an AI was trained legally; if the output are auditable, meaning whether can you decipher the derivation of their training material on a song by song basis; if there is a clear chain of training material ownership; what’s the protection and procedure for future rights disputes; and finally will the system comply with internal AI usage policies. 

If these questions can be answered successfully through the development of such an operational AI system and regulation, we’ll create a means for artist and rights holder remuneration that can ensure content creators and owners are properly paid for their contribution to generative AI.

So what does all this mean for you?

On a practical level, everything I mapped out will happen in some way, but it will take a while.  Until then, you need to make sure you keep informed on ongoing litigation and evolution of technology.  

And if you are working in sync, explore where you can use operational AI to help your music get discovered and delivered more successfully to opportunities, and be VERY CAREFUL about your usage of any generative AI systems in the creation of your music. 
Studios and networks currently ban the usage of AI-generated music on a blanket level, so if you are pitching music, steer clear from pitching anything that has been created in any manner or part by an AI music generator.  The stance of licensees will change over time, but for now, don’t pitch AI generated music.

And to keep you up to date on what’s happening in the industry around AI in all its forms, Sync Summit is launching a series of free, weekly online talks and will hold the LA edition of the AI and Sync Symposium at The Recording Club in Santa Monica and online on September 16 (sign up here now to attend online or in person) as well as publish an ongoing series of long-form articles providing breakdown and analysis of music and sync-related AI news.