AI Drake just set an impossible legal trap for Google

The AI Drake track that mysteriously went viral over the weekend is the start of a problem that will upend Google in one way or another — and it’s really not clear which way it will go.

Here’s the basics: there’s a new track called “Heart on My Sleeve” by a TikTok user called @ghostwriter877 with AI-generated vocals that sound like Drake and The Weeknd. The song mysteriously blew up out of nowhere over the weekend, which, well, is fishy for various reasons.

After the song went viral on TikTok, a full version was released on music streaming services like Apple Music and Spotify, and on YouTube. This prompted Drake and The Weeknd’s label Universal Music Group to issue a sternly-worded statement about the dangers of AI, which specifically says that using generative AI infringes its copyrights. Here’s that statement, from UMG senior vice president of communications James Murtagh-Hopkins:

UMG’s success has been, in part, due to embracing new technology and putting it to work for our artists–as we have been doing with our own innovation around AI for some time already. With that said, however, the training of generative AI using our artists’ music (which represents both a breach of our agreements and a violation of copyright law) as well as the availability of infringing content created with generative AI on DSPs, begs the question as to which side of history all stakeholders in the music ecosystem want to be on: the side of artists, fans and human creative expression, or on the side of deep fakes, fraud and denying artists their due compensation.

These instances demonstrate why platforms have a fundamental legal and ethical responsibility to prevent the use of their services in ways that harm artists. We’re encouraged by the engagement of our platform partners on these issues–as they recognize they need to be part of the solution.

What happened next is a bit mysterious. The track came down from streamers like Apple Music and Spotify which are in tight control of their libraries and can pull tracks for any reason, but it remained available on YouTube and TikTok, which are user-generated content platforms with established DMCA takedown processes. I am told by a single source familiar with the situation that UMG didn’t actually issue takedowns to the music streamers, and the streaming services so far haven’t said anything to the industry trade publications. Neither has Drake or The Weeknd. It’s weird – it does seem like Ghostwriter977 pulled the track themselves to create hype, especially while the song remained on YouTube and TikTok.

But then TikTok and YouTube also pulled the track. And YouTube, in particular, pulled it with a statement that it was removed due to a copyright notice from UMG. And this is where it gets fascinatingly weedsy and probably existentially difficult for Google: to issue a copyright takedown to YouTube, you need to have… a copyright on something. Since “Heart on my Sleeve” is an original song, UMG doesn’t own it — it’s not a copy of any song in the label’s catalog.

So what did UMG claim? I have been told that the label considers the Metro Boomin producer tag at the start of the song to be an unauthorized sample, and that the DMCA takedown notice was issued specifically about that sample and that sample alone. It is not clear if that tag is actually a sample or itself AI-generated, but YouTube, for its part, doesn’t seem to want to push the discussion much further.

“We removed the video after receiving a valid copyright notification for a sample included in the video,” is what YouTube spokesperson Jack Malon says about the situation. “Whether or not the video was generated using artificial intelligence does not impact our legal responsibility to provide a pathway for rightsholders to remove content that allegedly infringes their copyrighted expression.”

UMG has followed up by issuing individual URL-by-URL takedowns to YouTube as copies of the song pop up, all based on the Metro Boomin tag — I am told by another music industry source that the company can’t really use YouTube’s automated ContentID system because, again, it doesn’t own the song and can’t claim it for that system to begin matching it. (Oddly, Ghostwriter977 reuploaded the track to their YouTube page after the first takedown, and it’s… still there. Again, there’s a lot of fishy stuff going on here.)

Got all that? Okay, now here’s the problem: if Ghostwriter977 simply uploads “Heart on my Sleeve” without that Metro Boomin tag, they will kick off a copyright war that pits the future of Google against the future of YouTube in a potentially zero-sum way. Google will either have to kneecap all of its generative AI projects, including Bard and the future of search, or piss off major YouTube partners like Universal Music, Drake, and The Weeknd. Let’s walk through it.

The first legal problem with using AI to make a song with vocals that sound like they’re from Drake is that the final product isn’t a copy of anything. Copyright law is very much based on the idea of making copies — a sample is a copy, as is an interpolation of a melody. Music copyright in particular has been getting aggressively expansive in the streaming age, but it’s still all based on copies of actual songs. Fake Drake isn’t a copy of any song in the Drake catalog, so there’s just no dead-ahead copyright claim to make. There’s no copy.

Instead, UMG and Getty Images and publishers around the world are claiming that collecting all the training data for the AI is copyright infringement: that ingesting Drake’s entire catalog, or every Getty photo, or the contents of every Wall Street Journal article (or whatever) to train an AI to make more photos or Drake songs or news articles is unauthorized copying. That would make the fake Drake songs created by that AI unauthorized “derivative works,” and, phew, we’re still squarely in the realm of copyright law that everyone understands. (Or, well, pretends to understand.)

The problem is that Google, Microsoft, StabilityAI, and every other AI company are all claiming that those training copies are fair use — and by “fair” they do not mean “fair as determined by an argument in an internet comments section,” but “fair” as in “fair as determined by a court on a case-by-case application of 17 United States Code §107 which lays out a four-factor test for fair use that is as contentious and unpredictable as anything in American political life.”

I asked Microsoft CEO Satya Nadella about this when I talked to him about the new ChatGPT-powered Bing, and he wasn’t shy about it. “Look, at the end of the day, search is about fair use,” he said. “In other places, again, it’ll have to be really thought through as to what is the fair use. And then sometimes, I think there’ll be some legal cases that will also have to create precedent. “

That’s because there is no actual precedent for saying that scraping data to train an AI is fair use; all of these companies are relying on ancient internet law cases that allowed search engines and social media platforms to exist in the first place. It’s messy, and it feels like all of those decisions are up for grabs in what promises to be a decade of litigation.

So now imagine that you are Google, which on the one hand operates YouTube, and on the other hand is racing to build generative AI products like Bard, which is… trained by scraping tons of data from the internet under a permissive interpretation of fair use that will definitely get challenged in a wave of lawsuits. AI Drake comes along, and Universal Music Group, one of the largest labels in the world, releases a strongly worded statement about generative AI and how its streaming partners need to respect its copyrights and artists. What do you do?

• If Google agrees with Universal that AI-generated music is an impermissible derivative work based on the unauthorized copying of training data, and that YouTube should pull down songs that labels flag for sounding like their artists, it undercuts its own fair use argument for Bard and every other generative AI product it makes — it undercuts the future of the company itself.
• If Google disagrees with Universal and says AI-generated music should stay up because merely training an AI with existing works is fair use, it protects its own AI efforts and the future of the company, but probably triggers a bunch of future lawsuits from Universal and potentially other labels, and certainly risks losing access to Universal’s music on YouTube, which puts YouTube at risk.

I asked Google’s Malon about this dilemma, and he said “it’s not up to YouTube to determine who ‘owns the rights’ to content. This is between the parties involved, and is why we give copyright holders tools to make copyright claims and uploaders tools to dispute claims they believe are made incorrectly. Matters that cannot be resolved through our dispute process may ultimately need to be decided by a court.”

That’s the idea, but copyright claims on YouTube were messy and contentious before the AI explosion, and now there’s basically no way for Google to avoid some major litigation here.

YouTube only continues to exist because of a delicate dance that keeps rightsholders happy and the music industry paid, but the future of Google itself is a bet on an expansive interpretation of copyright law that every creative industry from music to movies to news hates and will fight to the death. Because it is death: generative AI tools promise to completely upend the market for almost all commodity creative work, and these companies are not required to sit back and let it happen.

The difference here is that the last time this sort of thing happened, Google and YouTube were disruptive upstarts with killer products and little to lose, and they accepted the litigation from Viacom and everyone else as the cost of winning. Now, they’re… well, YouTube is literally a cable company. And navigating the thorny world of content partnerships while chasing after AI startups who are much freer to break things will force Google to make almost impossible choices at every turn.