Suno & Udio: The AI-mpire Strikes Back!
These Silicon Valley Backed AI Music Generators Aren't Just Saying "Fair Use." They're Saying The Major Labels Are The Bad Guys.
Grab your coffee! It’s Monday morning brAIn dump time! First, VOTE below for my nominated generative AI/media roundtable discussion at SXSW 2025 (via the SXSW PanelPicker!). Next, it’s “the mAIn event” - my headline story about how music generative services are trying to turn the table on copyright owners that have claimed infringement. Finally, it’s “the AI case tracker” - updates on key generative AI copyright infringement cases.
And listen my “the brAIn” podcast on all major podcast platforms!
And please follow it, rate it, and share it. I record it myself, so you get my personal “flavor.” No AI/synthetic voices here!
I. VOTE NOW for My Nominated SXSW 2025 Panel
Please vote now for my nominated SXSW 2025 generative AI/media roundtable with expert Curt Doty via the SXSW PanelPicker (click this link or on the button below, or scan the QR code in the image below to vote). Our session is titled “Copy, Paste, Sue: The Tangled Web of AI and Copyright.” VOTE NOW to secure our spot!
II. the mAIn event - the Music Generative AI-mpire Strikes Back (& tries to turn the table on the major label copyright owners)
Early July, the major music labels sued music generative services Suno and Udio for alleged copyright infringement on a grand scale based on their “training” (scraping) of their copyrighted recordings without consent and compensation. I say “alleged” because at the time, both Suno and Udio — which are backed by the typical “who’s who” of Silicon Valley VC’s (including Andreessen Horowitz) — had vehemently resisted, time and time again, answering the pointed question whether they were, fact, training on copyrighted music. I had asked both companies myself to confirm or deny that allegation. And each time, I was met with crickets. All that these services would say — when drilled by the press — is that they trained their AI on “publicly available data” (those 3 words that serve as generative AI tech’s mantra).
Well, now we know — with 100% certainty — that training on “publicly available data” means mass scraping of copyrighted works (at least for them). And we know this because Suno and Udio both finally answered the question directly in their official court “Answers” to the major label lawsuits.
Suno wrote this in its court papers: “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 … Accordingly, Suno’s training data includes essentially all music files of reasonable quality that are accessible on the open Internet ….”
My initial reaction to Suno’s admissions — when I first read them — was “WOW!” On so many levels.
First, it’s a bit rich to write “It is no secret” that Suno trained its AI model on “essentially all files” on the Internet when Suno and Udio previously deflected time and time again. If it’s no secret, why were they so coy about it before?
So why did these two generative AI music companies now concede this fundamental question? Because they essentially had no choice in court, particularly in this case where the major label plaintiffs believe they found a proverbial “smoking gun.” One of Suno’s key early investors — in a momentary lapse of reason — said the quiet part out loud. Antonio Rodriguez of Matrix Partners said this: “Honestly if we had deals with labels when this company got started, I probably wouldn’t have invested in it. I think that they needed to make this product without the constraints.” Doesn’t sound like Antonio was thinking of “fair use” at the time.
“Constraints” — apparently that’s what content IP is to this VC (and to Suno and Udio by implication). That’s quite a remarkable statement when you think about it. Despite the fact that generative AI tech is of little value without the content required to train it (“data” in the parlance of big tech), many believe that valuable content can be just taken on grounds of “fair use.”
So what’s a generative AI company to do when faced with what the labels believe is a smoking gun? Try to turn the tables, that’s what. Suno’s and Udio’s lawyers — in a moment of Olympics-fueled inspiration — resort to legal gymnastics with a high degree of difficulty. It’s not Suno and Udio who are the bad guys here, they write. Rather, it’s the other way around. It’s the copyright owners themselves — the major record labels — that should be roundly chastised.
Labeling them as “brazen,” Suno and Udio’s lawyers write that the major labels “misuse IP rights to shield [themselves] from competition and reduce the universe of people who are equipped to create new expression.” In other words, those rights-holders — that have invested billions of dollars to build their content libraries — should be blamed for acting in concert to protect their copyrighted works.
It’s a clever “power to the people”, pro democratic (with a small “d”) type argument. Of course, Suno’s and Udio’s lawyers (both use the same firm, Latham & Watkins) overlook the fact that theirs is no philanthropic movement. These VC’s and execs are seeking to build commercial enterprises worth billions.
And even if something is good for society (and generative AI certainly holds remarkable promise), that doesn’t mean you can just take it. Several other generative music services accomplish the same laudable goal of giving the people what they want — but do it in a way that uses only licensed music for their AI training (in other words, an “ethically sourced” content approach).
Jen Music is one of them (I urge you to check it out). Jen Music proves that we can move forward with great generative AI innovation and — at the same time — properly seek consent and give fair compensation.
I ultimately believe we will develop a fair system of consent and compensation — that features an equitable allocation of value amongst copyright owners and generative AI tech — both on ethical and legal grounds (not to mention for the sheer expediency of removing the friction of risk that holds back progress). I’ve previously written about this many times (and refer you to my earlier legal analysis via this link where I discuss “fair use” arguments and analyze how I believe they will fail under the Supreme Court’s most recent major copyright infringement ruling in the “Andy Warhol-Prince” case).
What do you think? Send me your candid thoughts to peter@creativemedia.biz. I want to hear from you.
III. the AI legal case tracker - updates on key AI litigation
I lay out the facts - and the latest developments - via this link to the “AI case tracker” tab on “the brAIn” website. That’s where you’ll find everything you need about each of these cases (including my detailed analysis and predictions).
(1) The New York Times v. Microsoft & OpenAI
(2) UMG Recordings v. Suno
(3) UMG Recordings v. Uncharted Labs (d/b/a Udio)
(4) Universal Music Group, et al. v. Anthropic
(5) Sarah Silverman, et al. v. Meta (class action)
(6) Sarah Silverman v. OpenAI (class action)
(7) Getty Images v. Stability AI and Midjourney