UMG v. Anthropic AI "Spoiler Alert" - We've Seen This Movie Before

Big Tech Wins Again As It "Trains" Its AI On The Backs Of The Entertainment Industry & Copyrighted Works

Image Credit: TS2 Space

Last week, in an important development that impacts the entire entertainment industry, three major music publishers – Universal Music, Concord and ABKCO - sued Amazon-backed ChatGPT wannabe Anthropic for infringing their collective music lyrics on a massive scale. Amazon-backed Anthropic (which is investing up to $4 billion of our Prime money in the “startup”), among other things, trains its AI model by relentlessly “scraping” the entire web which means, of course, that it sucks massive numbers of copyrighted works into its vortex – all without any kind of licensing from creators. 

It’s a tale as old as time – an endless sequel of game-changing technology driving tech-tonic shifts in the media landscape that ultimately also drive major Hollywood/Silicon Valley litigation. Universal Music’s lawyers played a leading role in key litigation during tech’s earlier streaming revolution, so maybe we can take those lessons to anticipate what happens here for all sectors of media.

Let’s take our litigation time machine and go back a little less than 20 years ago to a simpler age largely devoid of video streaming and social media. That’s when a then-small little company known as YouTube rocked and roiled the entertainment industry seemingly out of nowhere, much like generative AI did less than one year ago (although it now feels like eons). YouTube, of course, breathlessly deployed technology that enabled anyone to “broadcast yourself” across the Internet. The problem was that the exuberant young upstart also enabled anyone to upload copyrighted works, such as clips of SNL’s iconic “Lazy Sunday” video that first sounded the alarm to Hollywood in late 2005 that we had entered an entirely new kind of tech driven media world order. 

Most people forget that at that same time, YouTube faced real competition from former Disney CEO Michael Eisner backed company Veoh that scaled massively itself. In fact, there is reason to believe that Veoh actually developed the ground-breaking concept and technology first (I was there at the time and have personal knowledge about it, but that’s another story). Universal Music took Veoh to court for enabling mass-scale copyright infringement, while Viacom took YouTube to court for its “brazen” actions - and those two pioneering (yet equally “guardrail-less”) tech-fueled Hollywood disruptors suffered entirely different fates. 

Universal Music sued Veoh into oblivion, whereas YouTube … well YouTube. YouTube withstood Viacom’s legal onslaught because, in the tech deal of the century, Google bought it for a mere $1.65 billion, financed YouTube’s defense, and bailed it out. The rest, as they say, is history.

Yes, due to the pressure and bad optics of litigation, Google was - shall we say - “incentivized” to develop and deploy new technology to reign in its otherwise unbridled copyright agnostic YouTube technology (it’s amazing what pressure, new priorities, and focused resourcing can do). That magical elixir is its “Content ID” system (first launched in 2007) that automatically detects whether uploaded content contains copyrighted material (music, television, film). If it does, and depending on the copyright owner’s wishes, YouTube either takes down the infringing content, or instead pays the man (or woman, as the case may be). It’s certainly far from perfect. But hey, it’s something.

Now let’s take this little tech-shocked entertainment industry history lesson back to the present day with Universal Music’s fresh lawsuit against Anthropic AI. Anthropic “feels” a lot more like YouTube than Veoh in this new scenario, since Amazon intends to fuel the company with its endless billions much like Google did with YouTube. That means that Anthropic’s coffers are rich enough to fight Universal and other big media players in the courts. 

And if Anthropic is more like YouTube than Veoh, then we can anticipate a similar fate here. Anthropic will cop to its past “misdeeds” and settle by paying an undisclosed sum of money that will appear to be massive to “insiders” at the time, but will be seen as being a trifle in the long-run. 

Universal Music’s litigation, together with the mountains of other related industry lawsuits in the courts right now, also will ultimately force Anthropic to either develop new tech that enables copyright holders to “opt in” to having their content scraped (which certainly doesn’t help them for all the unlicensed, non-consensual scraping by AI so far), or new forensic AI tech a la Content ID that leads to some new - and ultimately unsatisfying - royalty scheme for creators and copyright owners (Intel Labs for one is creating some AI forensic tech with its “My Art My Choice” initiative). Yes, the courts will hear Anthropic’s dire “fair use” pronouncements and draw some erratic and ambiguous lines about how much “scraping” of the works in question is too little to find infringement. But ultimately, Anthropic and the other major tech gorillAIs will have no choice. 

And if YouTube’s Content ID past is AI’s prologue, Silicon Valley’s giants will use their largely uncontrolled, copyrighted content-trained generative AI tech to generate accelerating billions (and ultimately likely trillions) to their bottom lines and overall market caps largely on the backs of creators. The end result, as we used to say in law school, is Q.E.D. – i.e., Latin for something akin to “the thing speaks for itself.” AI copyright litigation just becomes another cost of doing business for these tech behemoths. 

So how do artists, creators and copyright owners who now unwittingly train the AI tech in the first place make out (and even when it becomes wittingly)? Sure, many will learn to use generative AI as a tool to create cool new works (just like they learned to use YouTube). But let’s keep it real and look at today’s transformed industry economics and media company valuations for clues about who really wins. The WGA was essentially forced to punt the training issue down the road to end writer pain (i.e., the studios reserved their rights to “train” AI on pre-existing material), and actors continue to be on strike to set some basic guardrails. Meanwhile, mighty Disney, the fairest traditional media company of them all, sits at a nine year low $150 billion market cap, while tech giant Google/YouTube’s is nearly $2 trillion and Apple’s is nearly $3 trillion. 

You do the math. Big tech is the big winner here once again (and I’m no anti-tech guy; I’ve led several tech-forward media companies). This is no mere “Lazy Sunday” take-down situation where copyrighted works are easily identifiable. Here we have endless “Fake Drake” songs and deep fake videos, each capable of auto-spawning endless AI iterations, coming to a screen near you. And generative AI itself offers up few precise clues as to how it actually artificially creates. 

Just ask the CEOs of Google, Microsoft, and Chat GPT unleasher OpenAI who concede they are equally puzzled by how exactly their generative AI black boxes work, but have no intention of slowing down their collective gravy trAIn. And as for Anthropic, which markets itself as being the white knight amongst the others for creating “moral” and “ethical’ AI, oh yes, Google is an investor there too.

For those of you who push back and argue that humans “train” on pre-existing copyrighted works all the time when they create works inspired by (or even “in the style of”) of others, let’s be clear. They typically aren’t plagiarizing or making actual copies. But generative AI is when it “scrapes” each and every word.

NOTE: TODAY, October 25th at 9 am Pacific (12 noon Eastern), join me for my free “AI & Entertainment” webinar and Q&A. You can register here - and request my relevant deck at the time.