THE ECONOMIST: Record labels are in a spin over AI music as tech brings a new age of disruption

The music industry is no stranger to technological upheaval. From vinyl to cassettes to CDs to streaming, the recording and selling of music has been in a state of constant revolution.

The shift to streaming shook the industry hardest.

After the rise of illicit file-sharing in the Napster era, which eventually gave way to legal streaming, it took more than 20 years for revenue to return, even in nominal terms, to the level of the 1990s.

Generative artificial intelligence brings a new age of disruption. It threatens to transform the whole process of making music itself.

Music made by AI is already making its presence felt, in the zeitgeist as “slop” as well as in the charts. There is more dross than ever.

Alexis Lanternier, boss of Paris-based Deezer, says that about 90,000 fully machine-generated tunes a day are uploaded to his music-streaming service — half the total.

Spotify, the biggest streamer, says it has taken down 75 million “spammy tracks” in the past year.

Much of the slop is uploaded illegally by bot farms hoping to harvest fraudulent royalties. But many human artists use AI legitimately, to enhance their creativity.

“It’s like Ozempic,” says Mr Lanternier.

“Everybody is using it, but no one wants to say.”

A bigger question is the legality of AI-made music itself.

Some record labels accuse AI-music labs of piracy and have sued. Some have settled, hoping to make more money from licensing than from lawsuits. Confusingly, some do both.

As Mark Mulligan of MIDiA Research, a consultancy, puts it, all are grappling with the problem of how to stay in control of the industry when the ability to create music with AI lowers barriers to entry.

Many top music executives cut their teeth (and their first record contracts) in the Napster era. They do not want to get caught napping again.

The stock market suggests investors are worried. The share prices of the two biggest listed record labels, Universal Music Group (UMG) and Warner Music Group (WMG), have lagged behind the overall market over the past few years. (Sony Music, which ranks between them in revenue, is unlisted.)

The relationships to watch are those between all the big labels and Suno, a controversial four-year-old American AI-music company last valued at $5.4 billion (more than a third of WMG’s $14.2b market capitalisation).

This presents an intriguing natural experiment: UMG and Sony Music have sued Suno, doubling down with a case filed in September; WMG, having been part of the original suit, has settled.

In their new case UMG, Sony Music and others accuse Suno of ripping off more than 60,000 copyrighted recordings, by artists from Avril Lavigne to Justin Timberlake, and ingesting them into its models.

Under American copyright law damages can reach $150,000 per work, which may suggest the plaintiffs are pushing for a big payout (or exerting pressure to secure a juicy licensing deal).

Suno admits that the tens of millions of recordings used to train its models include copyrighted material. But it argues that it is “fair use” in law to learn musical styles and generate new, original music.

The matter is unsettled: those in the creative industries have more than 100 lawsuits in America against AI labs, including Anthropic and OpenAI, many of which turn on whether fair use applies.

While the Suno cases go through the courts, WMG’s settlement is also attracting attention.

The label has told investors that revenues from its AI partnerships will start flowing this financial year. They would be among the first AI deals in the entertainment industry to bring in real money.

Robert Kyncl, boss of WMG, says the firm’s decision to settle with Suno was based on expediency: it might take years for the courts to rule on fair use.

“If I can get a partner to give up their fair-use argument and start paying us, why would I wait?” he says.

The settlement itself may also herald big changes. In September, for instance, Suno introduced v6, which it says is its first family of models to be trained exclusively on licensed music from WMG and others.

Artists and songwriters who opt in will share revenues with Suno when their work is used, based on Suno’s subscriptions.

Mr Kyncl expects it to yield new revenue streams for artists, as fans create and post AI music derived from remixes or mashups of their repertoires.

As an example, he suggests Fleetwood Mac’s Dreams overlaid with Aretha Franklin’s voice.

“It just brings people joy.”

UMG has similar ambitions for fan remixes. This year it announced licensing deals with Spotify and ElevenLabs, an AI audio start-up, that would enable UMG’s material to be reworked by “superfans”.

Sam Duboff, Spotify’s head of marketing, says there is demand.

On some platforms, he reckons, unlicensed remixes and covers number one billion a month; Spotify wants to regularise the market.

Michael Nash of UMG says the company reckons 20-30 per cent of streaming customers would pay up to twice the price of their current subscription for access to AI-enabled superfan products.

A lot depends on how many leading performers opt in. Mr Nash hopes 1000 UMG artists will.

But so far UMG, WMG, Spotify and Suno have named no big stars who have signed up, suggesting that qualms related to artistic integrity remain high. And until top names agree to take part, it is hard to know how enthusiastic their fans will be.

MIDiA’s Mr Mulligan says partnerships between record labels and makers of AI music have a defensive quality: the industry’s giants, whether labels or streaming platforms, want to reinforce the status quo.

“They want to control AI, rather than have AI control them,” he said.

But they might be bypassed anyway. He believes the new technology will make it easier for new artists to make a splash without record deals, as well as enabling new streaming platforms to emerge.

“You end up with a more fragmented music world.”

Executives of UMG and Spotify disagree.

They bet that scarcity — the hallmark of the superstar era — will continue to prevail. (Taylor Swift, for example, is signed to UMG. )

UMG’s Mr Nash says the main beneficiaries of lower barriers to entry are not new artists, but “fraudsters who carpet-bomb (streaming) services with more content”.

Scarcity comes from clever marketing, promotion and audience development. If AI learns to replicate that, the music industry will be in deep trouble.