Spotify is now showing listeners when AI was used to make a song. But only when the artist says so.

The feature is called AI Credits, in beta, rolling out first through DistroKid. It sits inside Song Credits on mobile and flags AI involvement in specific parts of a track: vocals, lyrics, production. However, no credit doesn't mean AI wasn't used.

Apple Music launched something similar in March, called Transparency Tags, also self-reported. Deezer went a different direction and built detection, a tool that automatically flags fully AI-generated tracks, catching around 75,000 per day. Think about that number relative to what a voluntary disclosure system would cover.

Spotify has made real moves. More than 75 million spammy tracks removed in the past year, partnerships with Sony, Universal, Warner, Merlin, and Believe on what the company calls responsible AI products, and work through DDEX toward an industry-wide disclosure standard.

But infrastructure built on self-reporting depends on people using it honestly. A tag that only shows up when someone volunteers it tells you about the artists who wanted you to know. Detection is the only approach that doesn't start from that assumption. I think it has to be part of this, and I'm not sure the pace things are moving at reflects how fast the problem is growing.

Apple Music has never published its streaming numbers. In an industry where plays and listener counts get treated like quarterly earnings, that's a deliberate non-move.

Oliver Schusser, Vice President of Apple Music & International Content, talked through the reasoning recently. Apple's been in music for over fifty years. Logic, GarageBand, Shazam, iTunes, then streaming. They build the tools artists actually use to record. That's the frame everything else sits inside.

"We're doing this because we love music. We're doing this mostly to support artists."

Same logic on free tiers. Apple treats music like film or photography: something people spent real time making, worth paying for.

"Who are we to just give it away for free?"

The whole industry, in his view, would be in better shape without them.

It's a pointed thing to say when Spotify's user base is built almost entirely on free access. Apple isn't playing that game, which makes the position easier to hold.

I think he's right. Free tiers grew the user base but it also shrunk the check. Artists got paid from a smaller pool than they should have.

Spotify's Q1 numbers were good. 761 million monthly active users, up 12% year over year. Premium subscribers at 293 million. Revenue of €4.5 billion, operating income of €715 million, free cash flow of €824 million. All Q1 records. By any normal reading, this is a company doing well.

Shares dropped more than 10%.

The guidance was the problem. Spotify projected 299 million paid subscribers for Q2; analysts wanted 302 million. Operating income guidance came in at €630 million against expectations of roughly €684 million. Those aren't catastrophic misses. But markets had priced in more, and the reaction was swift.

The subscriber number matters to the music business in a specific way: the royalty pool tracks paid subscribers, not total users. Spotify says it paid out over $11 billion to the industry in 2025, up about 10% from the prior year. It also means subscriber growth slowing down isn't just a stock story, it eventually shows up in label revenue and artist payments.

On the product side, a few things shipped. Prompted Playlist expanded to Premium users in the U.S. and Canada, now including podcasts. Taste Profile launched in beta in New Zealand, letting users see and adjust what signals Spotify is using to build their recommendations. SongDNA went global in beta, giving listeners a way to explore who wrote, produced, or sampled a track. These are all about keeping existing users more engaged, none of them are really about acquiring new ones.

The Q2 forecast calls for 778 million total MAUs. That's 17 million new users in a quarter, which isn't a small number. The gap isn't user growth. It's that signing users up and getting them to pay are two different problems, and the second one is harder.

I don't think today's stock drop changes where Spotify sits in the music ecosystem. It's profitable, it's generating cash, and nobody else is close. But the free-to-paid conversion question has been sitting there for a while, and a record Q1 doesn't answer it.

Apple Music doubled its fraud penalties. The music industry's first criminal case just ended in a guilty plea. The pressure is coming from multiple directions now.

The first criminal conviction

Michael Smith made over $8 million streaming his own AI-generated songs to himself. Now he's facing up to five years in prison and has to give all of it back and more.

Smith admitted to generating hundreds of thousands of AI tracks and using over a thousand bot accounts to stream them billions of times. At its peak, his setup was pulling in roughly $3,300 a day. It ran for years across Apple Music, Spotify, Amazon Music, and YouTube Music before anyone caught it. This was the first criminal streaming fraud case in US history, and it ended in a guilty plea. Millions in royalties were pulled from artists and rights holders who actually earned them.

Apple's response

The same week, Apple Music announced it's doubling its fraud penalties. The old structure fined fraudsters between 5 and 25 percent of what they would have made. Now that's 10 to 50 percent. Commit $1 million in streaming fraud, and you're on the hook for an extra $500,000 on top of losing the original amount.

Apple's VP Oliver Schusser pointed to nearly 2 billion fraudulent streams caught and demonetized in 2025 alone, worth about $17 million in royalties that would have gone to people who didn't earn them. AI has made it cheap enough to flood a platform with throwaway content, and Apple's old penalty structure wasn't built for that volume.

The detection problem

The IFPI has been pushing for something more coordinated. Fraud detection only works if platforms, distributors, and aggregators share data. Right now, a bad actor removed from one platform can just move to another. Deezer reported that 60,000 AI songs are being uploaded to its platform every day, and 85 percent of the streams on those tracks are fraudulent. No single company's policy closes that gap.

A criminal conviction and a stiffer penalty structure in the same week isn't a coincidence. I think it's the beginning of real enforcement, and I'd expect more of both.

Suno just crossed $300 million in annual recurring revenue and 2 million paid subscribers. It got there in two years.

CEO Mikey Shulman posted the numbers on LinkedIn in late February, two years after the platform launched. Over 100 million people have used Suno since then.

Three months ago, the company reported $200 million in annual revenue when it closed a $250 million Series C at a $2.45 billion valuation. Going from $200 million to $300 million in roughly 90 days is fast for any company, let alone one that's still in active litigation with two major labels.

The platform generates about 7 million songs per day. Anyone can make one. You type a prompt, and Suno produces a finished track in seconds, no musical background required. A woman in Mississippi used it to turn her poetry into an R&B song that went viral and landed her a reported $3 million record deal. That's the range of what people are doing with it.

Warner Music Group settled with Suno in November and signed a licensing deal. Sony and Universal haven't. Both are still litigating, alleging Suno trained its models on copyrighted recordings without permission. European rights organizations are also pursuing claims. Earlier this week, leaders from the Music Artist Coalition, the European Composer and Songwriter Alliance, and the Artist Rights Institute published an open letter calling Suno a "brazen smash and grab."

At the same time, Suno has been quietly hiring from inside the industry. Former Merlin CEO Jeremy Sirota joined as Chief Commercial Officer. A former WMG executive is now Chief Music Officer. A former Spotify exec recently came on to lead artist partnerships. It's a strange position to be in: recruiting people whose former employers are suing you.

The unresolved question isn't whether Suno is growing. The question is what happens if Sony and Universal win. Udio, Suno's closest competitor, already settled with both. Suno hasn't, and whatever terms eventually get negotiated will probably set the standard for how AI music platforms license content going forward.

At $300 million in revenue and two years old, Suno isn't an experiment anymore. I think the licensing model wins eventually. The labels know they can't shut this down, and they'd rather get paid.

In January 2026, X filed an antitrust lawsuit against 18 music publishers and the NMPA, claiming publishers weaponized the DMCA's takedown process to pressure the platform into industry-wide licensing deals at inflated rates. Sony, Universal, and Warner are all named.

The NMPA started sending thousands of takedown notices per week in 2021. X's position: that wasn't copyright enforcement, it was a coordinated squeeze play. X says it's willing to negotiate directly with individual publishers. Publishers have refused, insisting on a blanket deal for the whole industry.

The publishers aren't buying that framing. NMPA president David Israelite called the lawsuit a "bad faith effort to distract from publishers' and songwriters' legitimate right to enforce against X's illegal use of their songs."

Every other major platform — YouTube, TikTok, Instagram, Twitch, Roblox, all of them — eventually signed licensing deals. X hasn't. When Musk took over, he walked away from talks Twitter had been in, reportedly deciding the fees were too high for a platform that doesn't center music. Publishers sued X in 2023 for up to $250 million over alleged infringement of 1,700 songs. Settlement talks got close last November, then collapsed.

That case goes to trial in February 2027. X's antitrust suit is a separate filing running alongside it.

The legal argument isn't without teeth; one antitrust expert told Bloomberg Law it has a pretty good chance of surviving a motion to dismiss. But surviving a motion to dismiss and winning are different things. Every platform that tried to wait out the music industry eventually paid. X is betting it can change the terms of how that happens. I'm not convinced the court is going to help it do that.

The example Söderström gave on the Q4 earnings call this week: an engineer opens Slack on their commute, tells Claude to fix a bug in the iOS app, and walks into the office to a waiting build. Not a demo. That's the actual workflow. The internal system is called "Honk," it runs through Claude Code, and senior developers, he said, haven't personally written a line since December.

Whether that's exciting or unsettling probably depends on where you sit. The output is real, Spotify shipped over 50 features in the past year, including AI Prompted Playlists, audiobook Page Match and About This Song. Söderström credited the pace to this workflow directly.                                                                              
Another interesting piece of the earnings call was what he said about data. Spotify's listening history captures something public datasets don't: where you're from shapes what you want to hear, and that specificity is baked into billions of sessions in ways that don't exist anywhere else. What counts as workout music in the U.S. looks different than it does in Scandinavia or Western Europe.That regional signal is what Spotify is training on.    

"This is a dataset that we are building right now that no one else is really building," Söderström said. "It does not exist at this scale."

On AI music: labels can now tag how much AI was involved in making a track, and Spotify is still actively removing what Söderström called "slop tracks."

The company is using AI heavily on the engineering side while trying to manage what it means for musicians on the supply side. Those two things will keep pulling against each other.          

The engineering shift has already happened. The actual question now is what the job is. If you're approving code you didn't write, engineer might not be the right word anymore. Editor, maybe. That's where I land.