The Journal

Suno & AI music, explained by people who use it daily.

We build Tha Mega Boy Band and teach AI:M Workshops on this stuff every week, so when something changes in Suno or the wider AI music world, here's our actual take — the good, the useful and the "read the fine print" moments.

UMG and Sony amend their Suno lawsuit with a new claim: cracking YouTube's encryption for training data

Universal Music Group and Sony filed an amended complaint against Suno on 25 August in Massachusetts federal court, adding a claim that Suno's training pipeline relied on stream-ripping tools such as yt-dlp to bypass YouTube's encryption and pull down copyrighted recordings wholesale. A judge had cleared the labels a week earlier to bring the claim under the DMCA's separate anti-circumvention provisions, on top of the standing copyright infringement claims.

This is a different flavour of accusation than the licensing fights we've been tracking all year, and it's the uglier one. Arguing you trained on recordings without a licence is one thing, courts are actively working out what that even means for AI. Arguing your team ran piracy tools against YouTube's copy protection to go get them is a much harder story to spin, because anti-circumvention claims don't hinge on the fuzzy "is training fair use" question at all, they're just about whether you broke the lock. We build our own catalogue on Suno every week and we're not walking that back over an allegation, but if discovery backs this up, "we scraped what we could reach" and "we cracked DRM on purpose" are two very different defences, and only one of them sounds remotely reasonable in front of a jury. Fact discovery closes 30 September, so we won't be waiting long to find out which one Suno actually has to run with.

Read the original report — Music Business Worldwide

Round Hill Music just sued Suno — and dragged Anthropic into court too

Independent music publisher Round Hill Music filed separate federal lawsuits against Suno and Anthropic on 17 August, alleging Suno trained on its master recordings and Anthropic trained Claude on its song lyrics, with damages in either case that could run past $1 billion. Round Hill says it isn't interested in a licensing settlement like the ones Suno struck with Warner Music and BMG, and wants a jury to decide instead.

What makes this one worth watching isn't the dollar figure, it's the split. Suno gets sued over the recordings, but Round Hill went after Anthropic separately over lyrics baked into a language model's training data, which is a reminder that the licensing mess we keep covering here isn't a Suno problem, it's a whole-AI-stack problem, and every company that trained first and asked questions later is going to get this call eventually. We run this business on both a language model and Suno, so we don't get to just watch this one from the sidelines. Round Hill skipping the settlement table for a jury trial is the real story: if that verdict lands against either company, "train now, license later" stops being a viable strategy for anyone in this space, not just Suno.

Read the original report — Variety

Suno signs BMG as its second major-label partner, and this time it's opt-in

Suno announced a global licensing agreement with BMG on 12 August, agreeing to pay for future use of BMG's recordings and cover the catalogue already folded into its past models, with only artists who actively opt in getting licensed at all. It's the second major label to settle and license with Suno, after Warner Music did the same earlier this year.

This is closer to the deal we've been hoping to see since we started building on Suno every week. Paying for what got used, and actually asking artists first instead of defaulting everyone in and leaving them to opt out after the fact, is the difference between "AI company that got caught" and "AI company artists can actually work with." If Suno holds that opt-in line as its next generation of models rolls out, it's a real answer to the "was a real artist even involved" question our AI:M Workshops students ask constantly. It's not the whole picture yet, though. UMG and Sony are still suing, so "industry-backed" currently means one out of three majors, and we won't know how strictly that opt-in actually gets enforced until the new models are out and someone goes looking for their voice in there without having agreed to it.

Read the original report — Music Business Worldwide

Suno is watermarking every song and putting downloads behind a paywall. Here's the trade-off

Suno CEO Mikey Shulman used a 6 August blog post, "How We're Building the Future of Music Responsibly," to announce that every track generated on the platform will now carry audio watermarking and fingerprinting, alongside a new downloads policy that requires a paid account and caps how many songs each tier can export per month. Updated community guidelines now explicitly ban faked vocals passed off as real people and the bot-driven streaming schemes that farm royalties out of mass-generated AI tracks.

We're glad to see this one land. The bot farms cranking out thousands of throwaway tracks to game Spotify royalties are exactly what gets "AI music" treated as a punchline before anyone's heard the actual work people are putting into it, ours included. Watermarking and fingerprinting are unglamorous, boring infrastructure, and that's exactly why it matters: it's the kind of groundwork that earns a platform credibility with the rest of the industry instead of just growth numbers. Where we'd push back a little is the download cap. Gating exports behind a paid plan with a monthly cieling is a blunt tool that catches working artists pulling stems and drafts right alongside the fraud accounts it's aimed at. If you're planning a release cycle in Suno, budget for that limit now rather than finding out mid-project.

Read the official announcement — Suno

A Munich court just ruled Suno's AI training is the copyright violation, not just what it spits out

Germany's GEMA won its case against Suno on 31 July, with the Munich Regional Court finding that Suno infringed copyright in six well-known songs, including "Rasputin", "Daddy Cool" and "Mambo No. 5", by training on them without a licence. Suno now has to open its books and disclose the revenue tied to that use, with damages to be set later, though it's entitled to appeal.

The part that actually matters here isn't the songs or the damages figure, which hasn't even been decided yet. It's that the court treated the training itself, done in the US, as the infringement once the model gets used from inside the EU. That's a bigger problem then a one-off payout, because it says where you trained doesn't get you out of where you're used. We build real releases on Suno and teach people to do the same in AI:M Workshops, so we're not cheering this on, but we're also not going to pretend it's nothing: it's GEMA's second win against an AI company in under a year, and it's the clearest signal yet that the licensing question isn't going away, no matter how good the model gets.

Read the original report — Variety
A person sitting alone at a laptop late at night, illustrating the Suno data breach story

Suno's data breach just exposed 55 million accounts. Here's what to actually do about it

Security researcher Troy Hunt's Have I Been Pwned service confirmed a data dump tied to Suno, affecting more than 55 million accounts. Mostly email addresses, with phone numbers thrown in for anyone who'd signed up that way.

If you've got a Suno account, don't just scroll past this one. Change your password, especially if you've reused it anywhere else (most of us have, no judgement). Turn on two-factor if it's offered, and be sceptical of any "Suno support" email landing in your inbox over the next few weeks, breaches like this are usually followed by a wave of phishing using the leaked addresses. We run our own catalogue through Suno too, so we're taking our own advise here.

Read the original report — The Register
A music industry professional reviewing something thoughtfully on a tablet in a bright studio, illustrating the AI music-labeling story

The music industry just rolled out AI labels, and honestly, we're for it

Eight major music organisations, including the RIAA, IFPI, the Recording Academy and SAG-AFTRA, announced a voluntary labelling system for streaming tracks: "AI-Generated" for songs where generative AI made most of the core elements, and "AI-Assisted" for tracks where a human did lead vocals or instruments with some AI help. Suno's response leaned on transparency and pointed to its existing investment in watermarking and audio fingerprinting.

As a studio putting out AI-composed music through Tha Mega Boy Band, we don't see this as a threat. If anything it's the opposite. A clear label doesn't undercut the case for AI music, it strenghtens it, because it removes the "wait, was a person actually involved?" question that follows every AI track around right now. If your process is genuinely collaborative, a label that says so is a good look, not a bad one.

Read the original report — Music Business Worldwide
A songwriter at a laptop late at night with a handwritten lyrics notebook beside him, illustrating the Suno lyrics editor story

Suno rebuilt its lyrics editor, and it fixes the thing everyone complained about

Suno shipped a redesigned lyrics editor with a dedicated Lyricist profile, natural-language editing (tell it what to change instead of hand-editing every line), rhyme and reference suggestions, and a distraction-free full-screen writing mode with autosave.

This one lands right on the workflow gap we talk about constantly in AI:M Workshops. Suno has always been great at turning a prompt into a finished track, but the old lyrics box always felt like an afterthought bolted onto a music tool. Autosave alone is going to save people alot of lost verses. If you tried writing your own lyrics in Suno once, gave up because the editor felt clunky, and let the AI generate them for you ever since, this update is worth a second look.

Read the official release notes — Suno
A musician reading a printed contract with a skeptical expression, illustrating the Suno Spark fine-print story

Suno's new artist grant program comes with a catch: don't ever say anything bad about Suno

Suno launched Spark, an incubator offering independent artists project funding, grants, marketing support and mentorship. Buried in the terms is a non-disparagement clause that bars participants, at any time, including after the program ends, from making any statement that puts Suno, its staff or its products in a negative light, on penalty of being removed from the program.

We'd genuinely be excited about this if it stopped at the grants and mentorship. More funding pathways for independent artists using AI tools is exactly the kind of thing our AI:M Workshops audience should have access too. But a non-disparagement clause that outlives the program itself is a hard pass in our book, and it's the same advice we'd give anyone in our workshops: read the terms before you sign, especially the clause that keeps applying after you've left. Free money that costs you your right to say what you actually think isn't free.

Read the original report — Stereogum
A team of entrepreneurs celebrating in front of a screen showing an upward-trending growth chart, illustrating the Suno funding and valuation story

Suno is now worth $5.4 billion, and users are making 7 million songs a day

Suno closed a $400 million round led by Menlo Ventures, pushing its valuation to $5.4 billion. At the time of the raise, users were making more than 7 million songs a day, and the app had hit #1 in the App Store's Music category in dozens of countries.

Seven million songs a day is a wild number, and it definately shows how mainstream AI music creation has gotten in a short window. We've been building on Suno since well before this raise, and the pace of improvement release to release backs that up. The open question is the right one though: a huge valuation doesn't settle whether the user base converts into durable, venture-scale revenue, and the copyright litigation working through the courts is still hanging over all of it.

Read the original report — Fortune
A vocalist singing passionately into a studio microphone with headphones on, illustrating the Suno v5.5 Voices story

Suno v5.5 landed with Voices, use your own vocals inside a generated track

Suno's v5.5 model launched with Voices, letting Pro and Premier subscribers capture and use their own singing voice inside generated tracks, with a verification step to confirm the voice is actually theirs. The release also added Custom Models trained on a subscriber's own catalogue, and My Taste, a personalisation feature available to everyone.

Voices is the update that made Suno feel less like "type a prompt, get a stranger's voice" and more like an actual instrument. For anyone building a real body of work, which is most of what we teach in AI:M Workshops, having a consistent voice across releases matters more then people expect going in. It's the difference between a one-off novelty track and something that sounds like an artist. If you tried Suno early and bounced off the generic vocals, this is the update that changes that.

Read the official announcement — Suno

Learn It Properly

Want to actually get good at this, not just read about it?

AI:M Workshops takes you from zero to shipping your own AI music and content — the same tools this Journal covers, hands-on.