UMG IS USING BLACKOPS AGAINST AI
In the final month of 2022, THE CADENCE wrote their first article about AI, speculating that the tech’s capacity for imitating our favorite songs (and the use of unapproved training data to enable such mimicry) would lead to a clash between rightsholders and developers as an army of machine learning musicians begins to infiltrate the DSPs.
Fast forward 100 or so days, and Universal Music Group has fired the first shots in this clone war that could define the future of music IP. The Financial Times is reporting that the world’s largest music company — which controls roughly a third of the recorded music market — has told DSPs, including Spotify and Apple Music, to block AI bots from scraping their catalog as it violates the copyrights owned by UMG and its artists.
The article also reports that UMG has been submitting takedown notices “left and right” against AI-generated music that has begun appearing on DSPs. It is not clear if those takedowns are justified by overt similarities to existing UMG IP or suspicion that said IP is being used in the training data. Regardless, it seems likely that the DSPs will acquiesce to the requests rather than take a stand that puts them firmly on the side of AI developers.
The question is, will it be enough? The majors have spent the past year complaining about the “flotsam and jetsam” and “lower-quality functional content” that has been slowly eating away at their share of revenues from the platforms. Until now, it was assumed that this content was mostly coming from some sort of royalty-free human source. AI has now thrown such assumptions into doubt.
Controlling the flow of data online is like trying to win a land war in Asia. That said, it’s wise of UMG to attack the supply lines that developers need to perfect their technology before the battle begins. These maneuvers even adds context to the extensive legal action taken by the RIAA against YouTube rippers in recent years — YouTube being a bottomless trove of licensed and unlicensed music that could be mined for AI training data.
Such precedent-setting could also make the DSPs culpable for any future damage caused by AI music if they fail to act today. There’s little doubt that the major label generals of 2023 are eager to avoid the mistakes of 2001, when the industry was slow to respond to the threat of online piracy while also failing to hold accountable the true profiteers of the Napster era — the broadband ISPs who raked in revenue while stolen IP bled freely across their networks for nearly a decade, only ending when the walled garden of the app store made legit streaming services the only practical option for mobile devices.
It’s also worth remembering that wars are not won by strategy and tactics alone. You must win hearts and minds — something the labels failed at miserably in the early-00s. That is why UMG has begun to wage a propaganda operation to bolster opinion in favor of human artists and villainize AI music. But not unlike Lars Ulrich talking to a scrum of reporters, the mission has been off to a clumsy start.
This became apparent when it was revealed that UMG failed to disclose its ownership of the semi-viral Human Artistry Campaign. Here, the obvious question is why UMG would choose to obscure its involvement in HAC? More concerning to some was the inclusion of the staunchly right-wing Center for Individual Freedom (CFIF) on the list of campaign supporters which also includes routinely progressive organizations like AFL-CIO.
The more important question is whether this matters. Existential threats can make strange bedfellows. Or, any enemy of AI is a friend of any group who profits from the copyright status quo. This includes institutional IP holders, labor unions, dark money lobbyists, professional sports player associations, performing rights organizations and independent music trade groups. What a party!
It’s reasonable to have a discourse over how the IP pie is divided. Just maybe not when the next generation of tech is threatening to swallow every last crumb.
Can it be stopped? UMG seems ready to try. But it will need to expand its coalition of the willing as well as its legal team. The U.S. Copyright Office has also indicated its willingness to try to hold back the AI onslaught by only allowing “AI-assisted” works to be eligible for protection. But will copyright even matter when the cost of content creation reaches zero?
And what, besides largely underdetermined case law about training AI on copyrighted materials that is slowly moving through the US and EU courts, is to keep Spotify from using its 70 million songs catalog to train its own music-making bots should investors start to demand profitability from the company? Daniel Ek’s recent tweet about ChatGPT and the “psychology of human misjudgment” was obtuse enough to be concerning.
