“You don’t need to train on novels and pop songs to get the benefits of AI in science” @ednewtonrex

You Don’t Need to Steal Art to Cure Cancer: Why Ed Newton-Rex Is Right About AI and Copyright

Ed Newton-Rex said the quiet truth out loud: you don’t need to scrape the world’s creative works to build AI that saves lives. Or even beat the Chinese Communist Party.

It’s a myth that AI “has to” ingest novels and pop lyrics to learn language. Models acquire syntax, semantics, and pragmatics from any large, diverse corpus of natural language. That includes transcribed speech, forums, technical manuals, government documents, Wikipedia, scientific papers, and licensed conversational data. Speech systems learn from audio–text pairs, not necessarily fiction; text models learn distributional patterns wherever language appears. Of course, literary works can enrich style, but they’re not necessary for competence: instruction tuning, dialogue data, and domain corpora yield fluent models without raiding copyrighted art. In short, creative literature is optional seasoning, not the core ingredient for teaching machines to “speak.”

Google’s new cancer-therapy paper proves the point. Their model wasn’t trained on novels, lyrics, or paintings. It was trained responsibly on scientific data. And yet it achieved real, measurable progress in biomedical research. That simple fact dismantles one of Silicon Valley’s most persistent myths: that copyright is somehow an obstacle to innovation.

You don’t need to train on Joni Mitchell to discover a new gene pathway. You don’t need to ingest John Coltrane to find a drug target. AI used for science can thrive within the guardrails of copyright because science itself already has its own open-data ecosystems—peer-reviewed, licensed, and transparent.

The companies like Anthropic and Meta insisting that “fair use” covers mass ingestion of stolen creative works aren’t curing diseases; they’re training entertainment engines. They’re ripping off artists’ livelihoods to make commercial chatbots, story generators, and synthetic-voice platforms designed to compete against the very creators whose works they exploited. That’s not innovation—it’s market capture through appropriation.

They do it for reasons old as time—they do it for the money.

The ethical divide is clear:

  • AI for discovery builds on licensed scientific data.
  • AI for mimicry plunders culture to sell imitation.

We should celebrate the first and regulate the second. Upholding copyright and requiring provenance disclosures doesn’t hinder progress—it restores integrity. The same society that applauds AI in medical breakthroughs can also insist that creative industries remain human-centered and law-abiding. Civil-military fusion doesn’t imply that there’s only two ingredients in the gumbo of life.

If Google can advance cancer research without stealing art, so can everyone else and so can Google keep different rules for the entertainment side of their business or investment portfolio. The choice isn’t between curing cancer and protecting artists—it’s between honesty and opportunism. The repeated whinging of AI labs about “because China” would be a lot more believable if they used their political influence to get the CCP to release Hong Kong activist Jimmy Lai from stir. We can join Jimmy and his amazingly brave son Sebastian and say “because China”, too. #FreeJimmyLai

by Trichordist

Sir Lucian Grainge Just Drew the Brightest Line Yet on AI

by Chris Castle

Universal Music Group’s CEO Sir Lucian Grainge has put the industry on notice in an internal memo to Universal employees: UMG will not license any AI model that uses an artist’s voice—or generates new songs incorporating an artist’s existing songs—without that artist’s consent. This isn’t just a slogan; it’s a licensing policy, an advocacy position, and a deal-making leverage all rolled into one. After the Sora 2 disaster, I have to believe that OpenAI is at the top of the list.

Here’s the memo:

Dear Colleagues,

I am writing today to update you on the progress that we are making on our efforts to take advantage of the developing commercial opportunities presented by Gen AI technology for the benefit of all our artists and songwriters.

I want to address three specific topics:

Responsible Gen AI company and product agreements; How our artists can participate; and What we are doing to encourage responsible AI public policies.

UMG is playing a pioneering role in fostering AI’s enormous potential. While our progress is significant, the speed at which this technology is developing makes it important that you are all continually updated on our efforts and well-versed on the strategy and approach.

The foundation of what we’re doing is the belief that together, we can foster a healthy commercial AI ecosystem in which artists, songwriters, music companies and technology companies can all flourish together.

NEW AGREEMENTS

To explore the varied opportunities and determine the best approaches, we have been working with AI developers to put their ideas to the test. In fact, we were the first company to enter into AI-related agreements with companies ranging from major platforms such as YouTube, TikTok and Meta to emerging entrepreneurs such as BandLab, Soundlabs, and more. Both creatively and commercially our portfolio of AI partnerships continues to expand.

Very recently, Universal Music Japan announced an agreement with KDDI, a leading Japanese telecommunications company, to develop new music experiences for fans and artists using Gen AI. And we are very actively engaged with nearly a dozen different companies on significant new products and service plans that hold promise for a dramatic expansion of the AI music landscape. Further, we’re seeing other related advancements. While just scratching the surface of AI’s enormous potential, Spotify’s recent integration with ChatGPT offers a pathway to move fluidly from query and discovery to enjoyment of music—and all within a monetized ecosystem.

HOW OUR ARTISTS CAN PARTICIPATE

Based on what we’ve done with our AI partners to date, and the new discussions that are underway, we can unequivocally say that AI has the potential to deliver creative tools that will enable us to connect our artists with their fans in new ways—and with advanced capability on a scale we’ve never encountered.

Further, I believe that Agentic AI, which dynamically employs complex reasoning and adaptation, has the potential to revolutionize how fans interact with and discover music.

I know that we will successfully navigate as well as seize these opportunities and that these new products could constitute a significant source of new future revenue for artists and songwriters.

We will be actively engaged in discussing all of these developments with the entire creative community.

While some of the biggest opportunities will require further exploration, we are excited by the compelling AI models we’re seeing emerge.

We will only consider advancing AI products based on models that are trained responsibly. That is why we have entered into agreements with AI developers such as ProRata and KLAY, among others, and are in discussions with numerous additional like-minded companies whose products provide accurate attribution and tools which empower and compensate artists—products that both protect music and enhance its monetization.

And to be clear—and this is very important—we will NOT license any model that uses an artist’s voice or generates new songs which incorporate an artist’s existing songs without their consent.

New AI products will be joined by many other similar ones that will soon be coming to market, and we have established teams throughout UMG that will be working with artists and their representatives to bring these opportunities directly to them.

RESPONSIBLE PUBLIC POLICIES COVERING AI

We remain acutely aware of the fact that large and powerful AI companies are pressuring governments around the world to legitimize the training of AI technology on copyrighted material without owner consent or compensation, among other proposals.

To be clear: all these misguided proposals amount to nothing more than the unauthorized (and, we believe, illegal) exploitation of the rights and property of creative artists.

In addition, we are acting in the marketplace to see our partners embrace responsible and ethical AI policies and we’re proud of the progress being made there. For example, having accurately predicted the rapid rise of AI “slop” on streaming platforms, in 2023 we introduced Artist-Centric principles to combat what is essentially platform pollution. Since then, many of our platform partners have made significant progress in putting in place measures to address the diversion of royalties, infringement and fraud—all to the benefit of the entire music ecosystem.

We commend our partners for taking action to address this urgent issue, consistent with our Artist-Centric approach. Further, we recently announced an agreement with SoundPatrol, a new company led by Stanford scientists that employs patented technology to protect artists’ work from unauthorized use in AI music generators.

We are confident that by displaying our willingness as a community to embrace those commercial AI models which value and enhance human artistry, we are demonstrating that market-based solutions promoting innovation are the answer.

LEADING THE WAY FORWARD

So, as we work to assure safeguards for artists, we will help lead the way forward, which is why we are exploring and finding innovative ways to use this revolutionary technology to create new commercial opportunities for artists and songwriters while simultaneously aiding and protecting human creativity.

I’m very excited about the products we’re seeing and what the future holds. I will update you all further on our progress.

Lucian

Mr. Grainge’s position reframes the conversation from “Can we scrape?” to “How do we get consent and compensate?” That shift matters because AI that clones voices or reconstitutes catalog works is not a neutral utility—it’s a market participant competing with human creators and the rights they rely on.

If everything is “transformative” then nothing is protected—and that guts not just copyright, but artists’ name–image–likeness (NIL), right of publicity and in some jurisdictions, moral rights. A scrape-first, justify-later posture erases ownership, antagonizes creators living and dead, and makes catalogs unpriceable. Why would Universal—or any other rightsholder—partner with a company that treats works and identity as free training fuel? What’s great about Lucian’s statement is he’s putting a flag in the ground: the industry leader will not do business with bad actors, regardless of the consequences.

What This Means in Practice

  1. Consent as the gate. Voice clones and “new songs” derived from existing songs require affirmative artist approval—full stop.
  2. Provenance as the standard. AI firms that want first-party deals must prove lawful ingestion, audited datasets, and enforceable guardrails against impersonation.
  3. Aligned incentives. Where consent exists, there’s room for discovery tools, creator utilities, and new revenue streams; where it doesn’t, there’s no deal.

Watermarks and “AI-generated” labels don’t cure false endorsement, right-of-publicity violations, or market substitution. Platforms that design, market, or profit from celebrity emulation without consent aren’t innovating—they’re externalizing legal and ethical risk onto artists.

Moral Rights: Why This Resonates Globally

Universal’s consent-first stance will resonate in moral-rights jurisdictions where authors and performers hold inalienable rights of attribution and integrity (e.g., France’s droit moral, Germany’s Urheberpersönlichkeitsrecht). AI voice clones and “sound-alike” outputs can misattribute authorship, distort a creator’s artistic identity, or subject their work to derogatory treatment—classic moral-rights harms. Because many countries recognize post-mortem moral rights and performers’ neighboring rights, the “no consent, no license” rule is not just good governance—it’s internationally compatible rights stewardship.

Industry Leadership vs. the “Opt-Out” Mirage

It is absolutely critical that the industry leader actively opposes the absurd “opt-out” gambit and other sleights of hand Big Technocrats are pushing to drive a Mack truck through so-called text-and-data-mining loopholes. Their playbook is simple: legitimize mass training on copyrighted works first, then dare creators to find buried settings or after-the-fact exclusions. That flips property rights on their head and is essentially a retroactive safe harbor,

As Mr. Grainge notes, large AI companies are pressuring governments to bless training on copyrighted material without owner consent or compensation. Those proposals amount to the unauthorized—and unlawful—exploitation of artists’ rights and property. By refusing to play along, Universal isn’t just protecting its catalog; it’s defending the baseline principle that creative labor isn’t scrapable.

Consent or Nothing

Let’s be honest: if AI labs were serious about licensing, we wouldn’t have come one narrow miss away from a U.S. state law AI moratorium triggered by their own overreach. That wasn’t just a safe harbor for copyright infringement, that was a safe harbor for everything from privacy, to consumer protection, to child exploitation, to everything. That’s why it died 99-1 in the Senate, but it was a close run thing,,

And realize, that’s exactly what they want when they are left to their own devices, so to speak. The “opt-out” mirage, the scraping euphemisms, and the rush to codify TDM loopholes all point the same direction—avoid consent and avoid compensation. Universal’s position is the necessary counterweight: consent-first, provenance-audited, revenue-sharing with artists and songwriters (and I would add nonfeatured artists and vocalists) or no deal. Anything less invites regulatory whiplash, a race-to-the-bottom for human creativity, and a permanent breach of trust with artists and their estates.

Reading between the lines, Mr. Grainge has identified AI as both a compelling opportunity and an existential crisis. Let’s see if the others come with him and stare down the bad guys.

And YouTube is monetizing Sora videos

[This post first appeared on Artist Rights Watch]

by Trichordist Editor

Artist Rights Are Innovation, Too! White House Opens AI Policy RFI and Artists Should Be Heard

The White House has opened a major Request for Information (RFI) on the future of artificial intelligence regulation — and anyone can submit a comment. That means you. This is not just another government exercise. It’s a real opportunity for creators, musicians, songwriters, and artists to make their voices heard in shaping the laws that will govern AI and its impact on culture for decades to come.

Too often, artists find out about these processes after the decisions are already made. This time, we don’t have to be left out. The comment period is open now, and you don’t need to be a lawyer or a lobbyist to participate — you just need to care about the future of your work and your rights. Remember—property rights are innovation, too, just ask Hernando de Soto (Mystery of Capital) or any honest economist.

Here are four key issues in the RFI that matter deeply to artists — and why your voice is critical on each:

1. Transparency and Provenance: Artists Deserve to Know When Their Work Is Used

One of the most important questions in the RFI asks how AI companies should document and disclose the creative works used to train their models. Right now, most platforms hide behind trade secrets and refuse to reveal what they ingested. For artists, that means you might never know if your songs, photographs, or writing were taken without permission — even if they now power billion-dollar AI products.

This RFI is a chance to demand real provenance requirements: records of what was used, when, and how. Without this transparency, artists cannot protect their rights or seek compensation. A strong public record of support for provenance could shape future rules and force platforms into accountability.

2. Derivative Works and AI Memory: Creativity Shouldn’t Be Stolen Twice

The RFI also raises a subtle but crucial issue: even if companies delete unauthorized copies of works from their training sets, the models still retain and exploit those works in their weights and “memory.” This internal use is itself a derivative work — and it should be treated as one under the law.

Artists should urge regulators to clarify that training outputs and model weights built from copyrighted material are not immune from copyright. This is essential to closing a dangerous loophole: without it, platforms can claim to “delete” your work while continuing to profit from its presence inside their AI systems.

3. Meaningful Opt-Out: Creators Must Control How Their Work Is Used

Another critical question is whether creators should have a clear, meaningful opt-out mechanism that prevents their work from being used in AI training or generation without permission. As Artist Rights Institute and many others have demonstrated, “Robots.txt” disclaimers buried in obscure places are not enough. Artists need a legally enforceable system—not another worthless DMCA-style notice and notice and notice and notice and notice and maybe takedown system that platforms must respect and that regulators can audit.

A robust opt-out system would restore agency to creators, giving them the ability to decide if, when, and how their work enters AI pipelines. It would also create pressure on companies to build legitimate licensing systems rather than relying on theft.

4. Anti-Piracy Rule: National Security Is Not a License to Steal

Finally, the RFI invites comment on how national priorities should shape AI development and it’s vital that artists speak clearly here. There must be a bright-line rule that training AI models on pirated content is never excused by national security or “public interest” arguments. This is a real thing—pirate libraries are clearly front and center in AI litigation which have largely turned into piracy cases because the AI lab “national champions” steal books and everything else.

If a private soldier stole a carton of milk from a chow hall, he’d likely lose his security clearance. Yet some AI companies have built entire models on stolen creative works and now argue that government contracts justify their conduct. That logic is backwards. A nation that excuses intellectual property theft in the name of “security” corrodes the rule of law and undermines the very innovation it claims to protect. On top of it, the truth of the case is that the man Zuckerberg is a thief, yet he is invited to dinner at the White House.

A clear anti-piracy rule would ensure that public-private partnerships in AI development follow the same legal and ethical standards we expect of every citizen — and that creators are not forced to subsidize government technology programs with uncompensated labor. Any “AI champion” who steals should lose or be denied a security clearance.

Your Voice Matters — Submit a Comment

The White House needs to hear directly from creators — not just from tech companies and trade associations. Comments from artists, songwriters, and creative professionals will help shape how regulators understand the stakes and set the boundaries.

You don’t need legal training to submit a comment. Speak from your own experience: how unauthorized use affects your work, why transparency matters, what a meaningful opt-out would look like, and why piracy can never be justified by national security.

👉 Submit your comment here before the October 27 deadline.

2025-18737Download
by thetrichordist

@DanMilmo: Top UK artists urge Starmer to protect their work on eve of Trump visit

UK artists including Paul McCartney, Kate Bush and Elton John urged Prime Minister Keir Starmer to protect creators before a UK-US tech pact tied to President Donald Trump’s visit. In a letter, they accuse Labour of blocking transparency rules that would force AI firms to disclose training data and warn proposals enabling training on copyrighted works without permission could let an artist’s life’s work be stolen. Citing human rights documents like the International Covenant on Economic, Social and Cultural Rights, the Berne convention and the European Convention on Human Rights, they frame the issue as a human-rights breach. Peer Beeban Kidron criticised US-heavy working groups. Government says no decision yet and promises a report by March. 

Read the post on The Guardian

by thetrichordist

Senator Josh @HawleyMO Throws Down on Big Tech’s Copyright Theft

 I believe Americans should have the ability to defend their human data, and their rights to that data, against the largest copyright theft in the history of the world. 

Millions of Americans have spent the past two decades speaking and engaging online. Many of you here today have online profiles and writings and creative productions that you care deeply about. And rightly so. It’s your work. It’s you.

What if I told you that AI models have already been trained on enough copyrighted works to fill the Library of Congress 22 times over? For me, that makes it very simple: We need a legal mechanism that allows Americans to freely defend those creations. I say let’s empower human beings by protecting the very human data they create. Assign property rights to specific forms of data, create legal liability for the companies who use that data and, finally, fully repeal Section 230. Open the courtroom doors. Let the people sue those who take their rights, including those who do it using AI.

Third, we must add sensible guardrails to the emergent AI economy and hold concentrated economic power to account. These giant companies have made no secret of their ambitions to radically reshape our economic life. So, we ought to require transparency and reporting each time they replace a working man with a machine.

And the government should inspect all of these frontier AI systems, so we can better understand what the tech titans plan to build and deploy. 

Ultimately, when it comes to guardrails, protecting our children should be our lodestar. You may have seen recently how Meta green-lit its own chatbots to have sensual conversations with children—yes, you heard me right. Meta’s own internal documents permitted lurid conversations that no parent would ever contemplate. And most tragically, ChatGPT recently encouraged a troubled teenager to commit suicide—even providing detailed instructions on how to do it.

We absolutely must require and enforce rigorous technical standards to bar inappropriate or harmful interactions with minors. And we should think seriously about age verification for chatbots and agents. We don’t let kids drive or drink or do a thousand other harmful things. The same standards should apply to AI.

Fourth and finally, while Congress gets its act together to do all of this, we can’t kneecap our state governments from moving first. Some of you may have seen that there was a major effort in Congress to ban states from regulating AI for 10 years—and a whole decade is an eternity when it comes to AI development and deployment. This terrible policy was nearly adopted in the reconciliation bill this summer, and it could have thrown out strong anti-porn and child online safety laws, to name a few. Think about that: conservatives out to destroy the very concept of federalism that they cherish … all in the name of Big Tech. Well, we killed it on the Senate floor. And we ought to make sure that bad idea stays dead.

We’ve faced technological disruption before—and we’ve acted to make technology serve us, the people. Powered flight changed travel forever, but you can’t land a plane on your driveway. Splitting the atom fundamentally changed our view of physics, but nobody expects to run a personal reactor in their basement. The internet completely recast communication and media, but YouTube will still take down your video if you violate a copyright. By the same token, we can—and we should—demand that AI empower Americans, not destroy their rights . . . or their jobs . . . or their lives.

by thetrichordist

Don’t forget tomorrow—Artist Rights Roundtable on AI and Copyright at American University in Washington DC

Artist Rights Roundtable on AI and Copyright: 
Coffee with Humans and the Machines     

Join the Artist Rights Institute (ARI) and American University’s Kogod’s Entertainment Business Program for a timely morning roundtable on AI and copyright from the artist’s perspective. We’ll explore how emerging artificial intelligence technologies challenge authorship, licensing, and the creative economy — and what courts, lawmakers, and creators are doing in response.

This roundtable is particularly timely because both the Bartz and Kadrey rulings expose gaps in author consent, provenance, and fair licensing, underscoring an urgent need for policy, identifiers, and enforceable frameworks to protect creators.

 🗓️ Date: September 18, 2025
🕗 Time: 8:00 a.m. – 12:00 noon
📍 Location: Butler Board Room, Bender Arena, American University, 4400 Massachusetts Ave NW, Washington D.C. 20016

🎟️ Admission: Free and open to the public. Registration required at Eventbrite. Seating is limited.

🅿️ Parking map is available here. Pay-As-You-Go parking is available in hourly or daily increments ($2/hour, or $16/day) using the pay stations in the elevator lobbies of Katzen Arts Center, East Campus Surface Lot, the Spring Valley Building, Washington College of Law, and the School of International Service

Hosted by the Artist Rights Institute & American University’s Kogod School of Business, Entertainment Business Program

🔹 Overview:

☕ Coffee served starting at 8:00 a.m.
🧠 Program begins at 8:50 a.m.
🕛 Concludes by 12:00 noon — you’ll be free to have lunch with your clone.

🗂️ Program:

8:00–8:50 a.m. – Registration and Coffee

8:50–9:00 a.m. – Introductory Remarks by KOGOD Dean David Marchick and ARI Director Chris Castle

9:00–10:00 a.m. – Topic 1: AI Provenance Is the Cornerstone of Legitimate AI Licensing:

Speakers:

  • Dr. Moiya McTier, Senior Advisor, Human Artistry Campaign
  • Ryan Lehnning, Assistant General Counsel, International at SoundExchange
  • The Chatbot

Moderator: Chris Castle, Artist Rights Institute

10:10–10:30 a.m. – Briefing: Current AI Litigation

  • Speaker: Kevin Madigan, Senior Vice President, Policy and Government Affairs, Copyright Alliance

10:30–11:30 a.m. – Topic 2: Ask the AI: Can Integrity and Innovation Survive Without Artist Consent?

Speakers:

  • Erin McAnally, Executive Director, Songwriters of North America
  • Jen Jacobsen, Executive Director, Artist Rights Alliance
  • Josh Hurvitz, Partner, NVG and Head of Advocacy for A2IM
  • Kevin Amer, Chief Legal Officer, The Authors Guild

Moderator: Linda Bloss-Baum, Director, Business and Entertainment Program, KOGOD School of Business

11:40–12:00 p.m. – Briefing: US and International AI Legislation

  • Speaker: George York, SVP, International Policy Recording Industry Association of America

🎟️ Admission:

Free and open to the public. Registration required at Eventbrite. Seating is limited.

🔗 Stay Updated:

Watch this space and visit Eventbrite for updates and speaker announcements.

by Trichordist

Why Artists Are Striking Spotify Over Daniel Ek’s AI-Offensive Weapons Bet—and Why It Matters for AI Deals

Over the summer, a growing group of artists began pulling their catalogs from Spotify—not over miserable and Dickensian-level royalties alone, but over Spotify CEO Daniel Ek’s vast investment in Helsing, a European weapons company.  Helsing builds AI-enabled offensive weapons systems that skirt international human rights law, specifically Article 36 of the Geneva Conventions. Deerhoof helped kick off the current wave; other artists (including Xiu Xiu, King Gizzard & the Lizard Wizard, Hotline TNT, The Mynabirds, WU LYF, Kadhja Bonet, and Young Widows) have followed or announced plans to do so.

What is Helsing—and what does it build?

Helsing is a Munich-based defense-tech firm founded in 2021. It began with AI software for perception, decision-support, and electronic warfare, and has expanded into hardware. The company markets the HX‑2 “AI strike drone,” described as a software‑defined loitering munition intended to engage artillery and armored targets at significant range—and kill people. It emphasizes resilience to electronic warfare, swarm/networked tactics via its Altra recon‑strike platform, and a human in/on the loop for critical decisions, and that limited role for humans in killing other humans is where it runs into Geneva Convention issues.   Trust me, they know this.

The X-2 Strike Drone

Beyond drones, Helsing provides AI electronic‑warfare upgrades for Germany’s Eurofighter EK (with Saab), and has been contracted to supply AI software for Europe’s Future Combat Air System (FCAS). Public briefings and reporting indicate an active role supporting Ukraine since 2022, and a growing UK footprint linked to defense modernization initiatives. In 2025, Ek’s investment firm led a major funding round that valued Helsing in the multibillion‑euro range alongside contracts in the UK, Germany, and Sweden.

So let’s be clear—Helsing is not making some super tourniquet or AI medical device that has a dual use in civilian and military applications.  This is Masters of War stuff.  Which, for Mr. Ek’s benefit, is a song.

Why artists care

For these artists, the issue isn’t abstract: they see a direct line between Spotify‑generated wealth and AI‑enabled lethality, especially as Helsing moves from software into weaponized autonomy at scale. That ethical conflict is why exit statements explicitly connect Dickensian streaming economics and streamshare thresholds to military investment choices.  In fact, it remains to be seen whether Spotify itself is using its AI products and the tech and data behind them for Helsing’s weapons applications.

How many artists have left?

There’s no official tally. Reporting describes a wave of departures and names specific acts. The list continues to evolve as more artists reassess their positions.

The financial impact—on Spotify vs. on artists

For Spotify, a handful of indie exits barely moves the needle. The reason is the pro‑rata or “streamshare” payout model: each rightsholder’s share is proportional to total streams, not a fixed per‑stream rate except if you’re “lucky” enough to get a “greater of” formula. Remove a small catalog and its share simply reallocates to others. For artists, leaving can be meaningful—some replace streams with direct sales (Bandcamp, vinyl, fan campaigns) and often report higher revenue per fan. But at platform scale, the macro‑economics barely budge.  

Of course because of Spotify’s tying relationships with talent buyers for venues (explicit or implicit) not being on Spotify can be the kiss of death for a new artist competing for a Wednesday night at a local venue when the venue checks your Spotify stats.

Why this is a cautionary tale for AI labs

Two practices make artist exits feel symbolically loud but structurally quiet—and they’re exactly what frontier AI should avoid:

1) Revenue‑share pools with opaque rules. Pro‑rata “streamshare” pushes smaller players toward zero; any exit just enriches whoever remains. AI platforms contemplating rev‑share training or retrieval deals should learn from this: user‑centric or usage‑metered deals with transparent accounting are more legible than giant, shifting pools.

2) NDA‑sealed terms. The streaming era normalized NDAs that bury rates and conditions. If AI deals copy that playbook—confidential blacklists, secret style‑prompt fees, unpublished audit rights—contributors will see protest as the only lever. Transparency beats backlash.

3) Weapons Related Use Cases for AI.  We all know that the frontier labs like Google, Amazon, Microsoft and others are all also competing like trained seals for contracts from the Department of War.  They use the same technology trained on culture ripped off from artists to kill people for money.

A clearer picture of Helsing’s products and customers

• HX‑2 AI Strike Drone: beyond‑line‑of‑sight strike profile, on‑board target re‑identification, EW‑resilient, swarm‑capable via Altra; multiple payload options; human in/on the loop.
• Eurofighter EK (Germany): with Saab, AI‑enabled electronic‑warfare upgrade for Luftwaffe Eurofighters oriented to SEAD/DEAD roles.
• FCAS AI Backbone (Europe): software/AI layer for the next‑generation air combat system under European procurement.
• UK footprint: framework contracting in the UK defense ecosystem, tied to strike/targeting modernization efforts.
• Ukraine: public reporting indicates delivery of strike drones; company statements reference activity supporting Ukraine since 2022.

The bigger cultural point

Whether you applaud or oppose war tech, the ethical through‑line in these protests is consistent: creators don’t want their work—or the wealth it generates—financing AI (especially autonomous) weaponry. Because the platform’s pro‑rata economics make individual exits financially quiet, the conflict migrates into public signaling and brand pressure.

What would a better model look like for AI?

• Opt‑in, auditable deals for creative inputs to AI models (training and RAG) with clear unit economics and published baseline terms.
• User‑centric or usage‑metered payouts (by contributor, by model, by retrieval) instead of a single, shifting revenue pool.
• Public registries and audit logs so participants can verify where money comes from and where it goes.
• No gag clauses on baseline rates or audit rights.

The strike against Spotify is about values as much as value. Ek’s bet on Helsing—drones, electronic warfare, autonomous weapons—makes those values impossible for some artists to ignore. Thanks to the pro‑rata royalty machine, the exits won’t dent Spotify’s bottom line—but they should warn AI platforms against repeating the same opaque rev‑shares and NDAs that leave creators feeling voiceless in streaming.

by thetrichordist

If you got one of these emails from Spotify, you might be interested

Spotify failed to consult any of the people who drive fans to the data abattoir: the musicians, artists, podcasters and authors.

Spotify has quietly tightened the screws on AI this summer—while simultaneously clarifying how it uses your data to power its own machine‑learning features. For artists, rightsholders, developers, and policy folks, the combination matters: Spotify is making it harder for outsiders to train models on Spotify data, even as it codifies its own first‑party uses like AI DJ and personalized playlists.

Spotify is drawing a bright line: no training models on Spotify; yes to Spotify training its own. If you’re an artist or developer, that means stronger contractual leverage against third‑party scrapers—but also a need to sharpen your own data‑governance and licensing posture. Expect other platforms in music and podcasting to follow suit—and for regulators to ask tougher questions about how platform ML features are audited, licensed, and accounted for.

Below is a plain‑English (hopefully) breakdown of what changed, what’s new or newly explicit, and the practical implications for different stakeholders.

Explicit ban on using Spotify to train AI models (third parties). 

Spotify’s User Guidelines now flatly prohibit “crawling” or “scraping” the service and, crucially, “using any part of the Services or Content to train a machine learning or AI model.” That’s a categorical no for bots and bulk data slurps. The Developer Policy mirrors this: apps using the Web API may not “use the Spotify Platform or any Spotify Content to train a machine learning or AI model.” In short: if your product ingests Spotify data, you’re in violation of the rules and risk enforcement and access revocation.

Spotify’s own AI/ML uses are clearer—and broad. 

The Privacy Policy (effective August 27, 2025) spells out that Spotify uses personal data to “develop and train” algorithmic and machine‑learning models to improve recommendations, build AI features (like AI DJ and AI playlists), and enforce rules. That legal basis is framed largely as Spotify’s “legitimate interests.” Translation: your usage, voice, and other data can feed Spotify’s own models.

The user content license is very broad. 

If you post “User Content” (messages, playlist titles, descriptions, images, comments, etc.), you grant Spotify a worldwide, sublicensable, transferable, royalty‑free, irrevocable license to reproduce, modify, create derivative works from, distribute, perform, and display that content in any medium. That’s standard platform drafting these days, but the scope—including derivative works—has AI‑era consequences for anything you upload to or create within Spotify’s ecosystem (e.g., playlist titles, cover images, comments).

Anti‑manipulation and anti‑automation rules are baked in. 

The User Guidelines and Developer Policy double down on bans against bots, artificial streaming, and traffic manipulation. If you’re building tools that touch the Spotify graph, treat “no automated collection, no metric‑gaming, no derived profiling” as table stakes—or risk enforcement, up to termination of access.

Data‑sharing signals to rightsholders continue. 

Spotify says it can provide pseudonymized listening data to rightsholders under existing deals. That’s not new, but in the ML context it underscores why parallel data flows to third parties are tightly controlled: Spotify wants to be the gateway for data, not the faucet you can plumb yourself.

What this means by role:

• Artists & labels: The AI‑training ban gives you a clear contractual hook against services that scrape Spotify to build recommenders, clones, or vocal/style models. Document violations (timestamps, IPs, payloads) and send notices citing the User Guidelines and Developer Policy. Meanwhile, assume your own usage and voice interactions can be used to improve Spotify’s models—something to consider for privacy reviews and internal policies.

• Publishers and collecting societies: The combination of “no third‑party training” + “first‑party ML training” is a policy trend to watch across platforms. It raises familiar questions about derivative data, model outputs, and whether platform machine learning features create new accounting categories—or require new audit rights—in future licenses.

• Policymakers: Read this as another brick in the “closed data/open model risk” wall. Platforms restrict external extraction while expanding internal model claims. That asymmetry will shape future debates over data‑access mandates, competition remedies, and model‑audit rights—especially where platform ML features may substitute for third‑party discovery tools.

Practical to‑dos

1) For rights owners: Add explicit “no platform‑sourced training” language in your vendor, distributor, or analytics contracts. Track and log known scrapers and third‑party tools that might be training off Spotify. Consider notice letters that cite the specific clauses.

2) For privacy and legal teams: Update DPIAs and data maps. Spotify’s Privacy Policy identifies “User Data,” “Usage Data,” “Voice Data,” “Message Data,” and more as inputs for ML features under legitimate interest. If you rely on Spotify data for compliance reports, make sure you’re only using permitted, properly aggregated outputs—not raw exports.

3) For users: I will be posting a guideline to how to clawback your data. I may not hit everything so always open to suggestions about whatever else that others spot.

Spotify’s terms give it very broad rights to collect, combine, and use your data (listening history, device/ads data, voice features, third-party signals) for personalization, ads, and product R&D. They also take a broad license to user content you upload (e.g., playlist art). 

Key cites

• User Guidelines: prohibition on scraping and on “using any part of the Services or Content to train a machine learning or AI model.”

• Developer Policy (effective May 15, 2025): “Do not use the Spotify Platform or any Spotify Content to train a machine learning or AI model…” Also bans analyzing Spotify content to create new/derived listenership metrics or user profiles for ad targeting.

• Privacy Policy (effective Aug. 27, 2025): Spotify uses personal data to “develop and train” ML models for recommendations, AI DJ/AI playlists, and rule‑enforcement, primarily under “legitimate interests.”

• Terms & Conditions of Use: very broad license to Spotify for any “User Content” you post, including the right to “create derivative works” and to use content by any means and media worldwide, irrevocably.

[A version of this post first appeared on MusicTechPolicy]

by thetrichordist

Waterloo Records remastered: Iconic vinyl shop celebrates grand re-opening

We could not be happier for our friends at Waterloo Records in Austin on their reopening down the street. Read about it here. Support your local record store!

The famous record shop, which for decades sat at the corner of West Sixth Street and North Lamar Boulevard, finally lowered the needle on its new location Saturday; the soundtrack a mashup of excited shoppers, intermittent announcements about prizes and giveaways, and, of course, music. 

Waterloo’s grand re-opening party marked the culmination of months of collaboration and planning among former majority owner John T. Kunz and new co-owners and operators Caren Kelleher and Trey Watson.

by Trichordist