Todd Edwards Samples Suno AI: What Producers Should Know

In an interview published by MusicRadar on 29 September 2026, house producer Todd Edwards said he now uses Suno Studio to generate material that he then chops and samples, and that the industry's attitude to AI is hypocritical. It matters because it lands in the middle of a major legal fight over AI music: a Munich court ruled against Suno on 31 July 2026, and Universal and Sony filed a second US lawsuit against the company on 18 September 2026. Meanwhile, the platforms underground labels depend on, from Bandcamp to Tidal, are drawing very different lines around what counts as acceptable AI music.

Who is Todd Edwards?

Todd Edwards is an American house and garage producer, DJ and singer, born in Bloomfield, New Jersey in 1972, and one of the people most closely associated with the sample-collage sound that fed into UK garage. According to his Wikipedia biography, he began producing in 1992, made his name in the 1990s on New York's Nervous Records, and broke through with remixes of St Germain's "Alabama Blues". His method is the thing people remember: hundreds of tiny vocal and instrumental fragments chopped, layered and rearranged until they read as new lyrics and new melodies over four-on-the-floor drums.

He is also a Daft Punk collaborator. He co-produced and sang on "Face to Face" from Discovery (2001) and co-wrote "Fragments of Time" on Random Access Memories (2013), which won Album of the Year at the 2014 Grammys. MusicRadar describes him as a "house legend" and UKG innovator with a 34-year career, and credits him as an influence on acts from Burial to Disclosure. That biography is why this interview carries weight. Edwards is not a newcomer chasing a shortcut. He is a producer whose whole style grew out of working with other people's recordings.

What did Todd Edwards say about AI?

The interview, written by Ana Yglesias, is mostly a gear and history conversation: Ensoniq and Akai samplers, a Juno-106, the limits of early hardware. The AI section arrives late. MusicRadar reports that Edwards says he felt like he was "going through the motions" at the start of 2025, making his own samples with synths and Splice, which he found depressing. Using Suno Studio to generate material he can safely sample, he says, reignited his interest in sample-based work.

Three claims stand out.

It is a tool for experienced people. Edwards says the seasoned producer sees AI as a powerful tool while the amateur sees a threat. In the same interview he says he rejects the framing of the technology as evil: "I resent the way people talk about this technology as if it's from the devil."

The sample-pack comparison. He argues that building from AI-generated stems is no different from building from Splice loops, and that critics who say AI takes money from artists ignore how much uncleared sampling already happens in dance music. He asks whether every artist on Beatport clears their samples, and answers that if not, "it's the same thing, just on a smaller scale". His line on this is that "The hypocrisy in the industry is insane."

Ownership. He says he makes his own disco or soft rock songs in Suno, cuts them up, and uses them as samples without worrying about "another lawsuit again". He adds that prompting alone gives generic results and that putting in his own ideas makes it act "like a co-producer". MusicTech's summary of the interview, published on 2 October 2026, picks out the same themes.

One piece of context belongs to MusicRadar rather than Edwards: the outlet notes that AI makes uncleared samples easier to identify and that copyright suits are increasingly common, which is part of why he changed approach. Edwards is also careful to say he does not need AI to make records. He says he could lose his sampler and still make house and garage.

What is Suno Studio?

Suno is the company behind a text-to-music generator. Suno Studio is its attempt at something closer to a production environment. In its launch announcement dated 25 September 2025, Suno called it a generative audio workstation, available at launch to Premier subscribers on desktop. The post describes generating stem variations for parts such as vocals, drums and synths, multitrack timeline editing, tempo and pitch controls, starting from uploaded samples, your Suno library or individual stems, and exporting stems as both audio and MIDI to use in another DAW.

That feature list explains Edwards' workflow. He is not publishing finished Suno songs. He is using the generator as a source of raw material, then doing what he has always done: cutting it into something with his own fingerprints.

On rights, Suno's own help page (last edited 7 January 2026) says users on the Pro or Premier plans are considered owners of songs created while subscribed, and that free-tier songs remain Suno's, usable for non-commercial purposes only. Music in Africa reported in December 2025 that Suno was revising its ownership wording after the Warner deal, so anyone relying on this should read the current terms rather than a summary.

Is AI sampling legal in 2026?

Nobody can give a clean yes or no, and this article is not legal advice. What can be said is what courts and companies have actually done. No court has declared Edwards' workflow unlawful, and the lawsuits below are against Suno, not its users. But the questions underneath it are open.

What are the courts doing about Suno?

The labels sued Suno in the US in June 2024. In November 2025, Warner Music Group settled and signed a licensing deal with the company, reported by Complete Music Update on 26 November 2025. Under it, downloads require a paid account, artists are asked for consent before their voice or likeness is used, and a licensed service was expected in 2026.

Universal and Sony did not settle. As Music Business Worldwide reported, they filed a second suit in Boston on 18 September 2026 covering 60,202 recordings, arguing that Suno's v6 models, launched on 9 September 2026, descend from models trained on their recordings without a licence. The complaint calls v6 "the fruit of the same poisoned tree". Suno's chief product officer has said v6 was trained from scratch on data that excludes Universal and Sony material, and the company has called training on recordings "quintessential fair use" in the earlier case. Both sides' positions are still just positions.

Outside the US, the picture is harsher for Suno. On 31 July 2026 the Munich Regional Court I ruled largely for the German collecting society GEMA in case 42 O 763/25, over six well-known compositions. As JUVE Patent reported, the court ordered Suno to stop reproducing the works and using them for training, to disclose related revenue, and to pay damages in an amount still to be set. The judgment is not final and Suno said it would explore all options, including an appeal.

One more data point from the same week as the Edwards interview: the Third Circuit affirmed on 29 September 2026 that Ross Intelligence's use of Thomson Reuters headnotes to train a non-generative tool was not fair use, according to Digital Music News. That ruling is narrow and does not decide the generative cases, but it shows courts weighing market harm seriously.

Who owns what a model generates?

Separate from training is the output. In its January 2025 report on copyrightability, the US Copyright Office concluded that, with current technology, prompts alone do not give users enough control to be the authors of what a system produces. It also said human authors keep copyright in their own expression that is perceptible in AI outputs, and in their creative selection, arrangement and modification of those outputs. A contract that calls you the owner is not the same thing as copyright protection.

What does sample clearance history tell us?

Sampling law has never been settled by one tidy ruling. In 1991, Grand Upright v. Warner Bros. held Biz Markie's use of Gilbert O'Sullivan's "Alone Again (Naturally)" to be infringement, with the judge opening his opinion with "Thou shalt not steal". In 2005 the Sixth Circuit's Bridgeport v. Dimension Films went further, saying "Get a license or do not sample" and refusing a de minimis defence for sound recordings. In 2016 the Ninth Circuit's VMG Salsoul v. Ciccone reached the opposite view on a 0.23-second horn hit in Madonna's "Vogue", holding that de minimis applies. Two circuits, two answers.

That history cuts both ways for Edwards. It supports his claim that clearing hundreds of micro-samples is a Herculean task, which is MusicRadar's description of his catalogue. It also shows that "everyone does it" has never been a legal defence.

How does this differ from the Spotify, Deezer and Bandcamp positions?

Edwards is talking about a producer using generated material as an ingredient. The platforms are mostly talking about something else: generated finished tracks, or fake artists, uploaded in volume. The table summarises where each stands, drawing on our earlier coverage and, for Tidal, Mixdown's report of 30 June 2026.

Read as a set, the policies target output type and artist identity, not workflow. A track in which a producer has chopped, resequenced and arranged generated stems is not "wholly" AI-generated in Tidal's terms and is not a fake persona in Spotify's. Bandcamp is the unclear case: "substantial part" is a judgement call, and as our six-month review found, enforcement has run on suspicion rather than a published standard.

What does it mean for underground producers?

For most working electronic producers, the immediate question is not whether to copy Edwards but how to stay out of trouble and keep their credibility. Our look at the real numbers behind being an underground DJ shows how thin the margins already are, citing GigXchange data that puts typical 2026 UK fees at GBP 100 to GBP 400 for a standard evening slot. In that economy, a takedown or a rights dispute costs more than any production shortcut saves.

There is also a distinction worth keeping. AI that takes apart existing human recordings, as covered in our piece on AI stem separation for DJs, is a different thing from generation. Edwards' workflow sits between the two: it generates new material, but his stated purpose is to replace sample sources, not to replace the producer.

What we make of it

What follows is RTS.FM's own opinion, not reporting. Each point comes with the strongest counter-argument we can find.

1. A tool in a producer's hands is not the same problem as an upload flood

The harm that platforms and prosecutors have documented comes from volume. In the Michael Smith case, a man who uploaded AI songs and ran bots against them, at one point about 661,440 streams a day, was sentenced on 6 October 2026 to 18 months and ordered to forfeit $8,091,843.64. Pro-rata payment means every fake stream shrinks everyone else's share. A producer who turns a few generated stems into one finished record, with human arrangement and a name on it, does not touch that mechanism.

The counter-argument: the tool and the flood run on the same models, and intent is invisible in the output. Deezer says AI-generated tracks now arrive at roughly 90,000 a day, and every generation request, careful or not, runs on the same models. Individual restraint does not change the market the tool feeds.

2. The Splice comparison holds in craft and fails in provenance

Edwards is right that producers have long built tracks from other people's sounds, and that nobody calls a record fake because it uses a loop pack. Creatively, the comparison is fair: selection, chopping and arrangement are where authorship lives, and the Copyright Office's own line on human selection and arrangement supports it. But the legal and ethical structure differs. Splice says its sounds are royalty-free and cleared for commercial use under its terms, which is a documented chain from contributor to user. Suno's chain is contested in court by Universal and Sony, Suno lost a first-instance case in Munich, and only part of the industry has licensed it, through Warner. A pack licence tells you who agreed to what. A generator currently cannot.

The counter-argument: Edwards' point about Beatport stands as an observation. Uncleared samples have circulated in dance music for decades, enforcement has always been uneven, and Warner's deal suggests a licensed route is forming, which could close the gap between Splice and Suno within a year or two.

3. Underground labels need a provenance norm before platforms impose one

If the policies above judge output type, the sensible response for small labels is boring and practical: keep session files and stems, record which sounds came from where, and be ready to say plainly when a track contains generated material. We argue for a simple credit line and a paper trail as a norm, not a rule. It protects an artist if a release is questioned, which our Bandcamp review showed can happen with little warning, and it keeps the word "human-made" meaningful for listeners.

The counter-argument: disclosure can backfire. Platform labels target personas and fully generated tracks, so volunteering a credit invites suspicion without earning any protection, and Bandcamp's suspicion-based enforcement punishes visible AI use more than hidden use. Honest producers would pay a reputational cost that careless ones avoid.

4. "Legally safe" is a marketing phrase, not a finding

Edwards says the material is "mine, all of it". That may feel true in the studio, but it is not what the law currently says. Copyright attaches to human contribution, not to prompts alone. A subscriber's contractual ownership under Suno's terms does not bind a court or a plaintiff. And the training questions, from Boston to Munich, are still being argued, with first-instance rulings that can be appealed. "Safer than sampling a record you cannot clear" is a defensible claim. "Never worry about another lawsuit" is not one we would repeat.

The counter-argument: for a producer who transforms generated material beyond recognition, the practical risk may be very small, smaller than the risk in the uncleared samples Edwards is comparing it to. No court that we found has pursued an end user, and perfect legal certainty has never been available in sampling.

Frequently Asked Questions

Who is Todd Edwards?

Todd Edwards is an American house and garage producer and DJ, born in New Jersey in 1972, known for chopping vocal and instrumental samples into collages. He co-produced Daft Punk's "Face to Face" on Discovery (2001) and co-wrote "Fragments of Time" on Random Access Memories (2013).

What did Todd Edwards say about AI?

In a MusicRadar interview published on 29 September 2026, he said he uses Suno Studio to generate material to sample, that the technology is a powerful tool for experienced producers, and that the industry's attitude to AI is hypocritical.

What is Suno Studio?

Suno Studio is a generative audio workstation launched by Suno on 25 September 2025. It lets users generate stem variations, edit them on a multitrack timeline, and export stems as audio and MIDI. It launched for Premier subscribers.

Is using AI-generated samples legal in 2026?

No court has declared a producer's use of AI-generated samples unlawful, but the underlying questions are unresolved. Suno faces ongoing lawsuits from Universal and Sony in the US and lost a first-instance case to GEMA in Munich on 31 July 2026. This is not legal advice.

Has Suno settled its lawsuits with the major labels?

Only partly. Warner Music Group settled and signed a licensing deal with Suno in November 2025. Universal and Sony are still litigating and filed a second suit on 18 September 2026.

Can you copyright music made with Suno?

According to the US Copyright Office's January 2025 report, prompts alone do not make a user the author of AI output. Human contributions such as original expression, creative selection, arrangement and modification can be protected.

Does Bandcamp allow AI-generated music?

No. Bandcamp bans music generated wholly or in substantial part by AI, enforced through human review and user reports. Its policy was announced in January 2026.

How does Tidal treat AI-generated music?

From 15 July 2026, Tidal labels tracks it identifies as wholly AI-generated and pays no royalties on them. It still accepts such music on the platform and removes AI tracks tied to fraud or impersonation.

Is a Splice loop the same as an AI-generated sample?

Creatively they are similar, since both are raw material a producer arranges. Legally they differ: Splice says its sounds are royalty-free and cleared for commercial use under its terms, while the rights questions around Suno's training data are still being argued in court.

What should underground labels do about AI-assisted tracks?

There is no official standard yet. A practical approach is to keep session files and stems, document where each sound came from, and check each platform's current AI policy before releasing.