Deezer has built a powerful public narrative around protecting artists from the worst excesses of modern streaming. It detects AI-generated music, identifies suspicious streams, removes fraudulent activity from royalty calculations, excludes fully AI-generated tracks from recommendations and now sells its detection technology to other organisations across the music industry. On the surface, this looks like exactly what musicians should want from a streaming company. The problem begins when the company asking artists to trust its systems also becomes the company defining the categories, operating the detection technology, deciding what content receives recommendation exposure and determining which streams are legitimate.
Deezer says its AI detection system has 99.8% accuracy and fewer than one false positive per 10,000 genuine human-made tracks according to its own measurements, while acknowledging that the system can still miss AI-generated tracks. That is a remarkable technical claim, but it should not end the conversation. It should start it. Because an independent artist should never have to confuse a company’s confidence in its own algorithm with accountability to the people whose livelihoods are affected by that algorithm.
Deezer has correctly identified a serious problem. The platform says AI-generated music represented more than half of its daily uploads at peak levels in June 2026, with roughly 90,000 AI-generated tracks arriving each day. It also says that up to 85% of streams associated with fully AI-generated music were fraudulent in 2025, while fraudulent streams across its wider catalogue represented about 8% of all streams. Those figures expose a genuinely ugly side of the streaming economy: it is now possible to manufacture music at industrial scale, distribute it cheaply and combine enormous catalogues with artificial listening in an attempt to extract money from a royalty system designed around human consumption.
Deezer deserves credit for detecting and publicising that problem. But there is a dangerous leap between saying “there is fraud” and saying “therefore the platform should be trusted to decide what is legitimate.” Those are two entirely different propositions. The first is an observable industry problem. The second is a question of corporate power.
The old idea of a streaming service was relatively simple. A company licensed music, delivered it to listeners and collected subscription or advertising revenue. Modern streaming is considerably more complicated. Platforms now determine discovery, personalise recommendations, analyse listening behaviour, police fraudulent activity, classify recordings, influence visibility and increasingly develop technologies that can be licensed back into the wider music industry. Deezer’s own AI programme demonstrates that transformation.
It began deploying its detection technology in early 2025, introduced explicit AI tagging later that year and made the technology commercially available to other music-industry organisations in January 2026. That progression matters because Deezer is no longer simply participating in the music market. It is building infrastructure that can influence how the market understands music.
That should make every independent artist uncomfortable.
Deezer says fully AI-generated tracks are automatically removed from algorithmic recommendations and editorial playlists. The company frames this as a protective measure intended to stop synthetic catalogue from diluting the royalty pool and overwhelming genuine music. There is an obvious argument for doing this when dealing with industrial-scale synthetic uploads designed to generate fraudulent streams. But artists need to understand the larger principle: a classification is not merely information if it changes how the platform distributes attention. The moment an AI label affects recommendation eligibility, it becomes economically meaningful. Discovery determines streams. Streams influence visibility.
Visibility influences audience growth. Audience growth influences future income. A supposedly informational label can therefore become part of a chain of commercial consequences. That is why the debate cannot stop at whether Deezer’s detector works. The real question is whether artists have sufficient transparency and recourse when the detector’s output changes what happens to their music.
Deezer describes its detection technology in impressive technical terms. It says the system identifies inaudible signatures associated with generative AI models and can identify models such as Suno and Udio. The company has also filed patent applications covering detection techniques designed to distinguish synthetic material from authentic recordings. But the artist confronted by a classification does not receive the source code, the detector’s complete evidentiary record or an independent audit of the decision. That asymmetry is the heart of the accountability problem. Deezer knows how its system operates.
Deezer owns the infrastructure. Deezer controls the implementation. The artist is expected to live with the outcome. In almost every other area of serious professional life, a decision capable of affecting someone’s economic interests would invite questions about evidence, process and appeal. In digital music, those protections can become much harder to see.
The number sounds almost definitive because it is presented as one. But even Deezer’s own description makes clear that the system is not perfect: the company says it may miss approximately two AI-generated tracks in every 1,000 and may incorrectly flag fewer than one genuine human-made track in every 10,000. That means Deezer itself recognises that errors exist. The important question is therefore not whether the system is accurate in aggregate. It is what happens to the individual artist who falls into the error category.
A statistical model can perform brilliantly across millions of tracks while still creating an unacceptable experience for the particular musician whose album is misclassified. The individual does not receive 99.8% of a career. The individual receives one decision.
The deeper problem is philosophical. Audio can be measured. Creativity cannot be reduced so easily. A recording may contain digital instruments, synthesis, sampling, machine-assisted mastering, pitch correction, automated editing, algorithmic effects and other technologies without becoming equivalent to a fully machine-generated work. Deezer’s public position specifically focuses its tagging and recommendations policy on fully AI-generated content, which is an important distinction. But the closer the industry moves toward automated classification, the more important those distinctions become.
Independent musicians are among the people most likely to experiment with unconventional production methods because they often lack the rigid workflows of major commercial studios. The very qualities that make underground music interesting can also make it difficult to categorise neatly. A platform that treats classification as an administrative exercise risks forgetting that experimental music has always existed precisely because artists refuse neat categories.
A major superstar has lawyers, management, label representatives, distributor contacts and enough commercial importance to make a platform respond quickly. A small independent artist may have none of those things. That difference is crucial. When a platform makes a mistake, the cost of challenging that mistake is not distributed equally. An established artist can escalate. An unknown artist can spend days sending support messages and still struggle to find a human being who understands the problem. A delay that is irritating for a superstar can be devastating for an independent musician trying to build momentum around a release. This is why platform accountability matters more for small artists than for the biggest names. The people with the least leverage need the strongest procedural protections.
Deezer has also promoted its artist-centric payment model as a way to improve remuneration. Under its model, certain artists meeting thresholds for genuine engagement can receive a higher payment multiplier, while tracks identified as noise or fraudulent activity can be excluded from the royalty pool. Deezer and SACEM announced an artist-centric payment system for publishing rights in France in 2025, including a two-times multiplier for songs meeting specified listener-engagement criteria. These reforms deserve serious examination rather than automatic dismissal. Changing the economics of streaming is difficult, and Deezer has at least attempted to address problems embedded in the traditional model.
But a better payment formula does not eliminate the power imbalance between the platform and the creator. You can redesign royalties while still leaving artists dependent on opaque systems for discovery, classification and enforcement.
Fairer economics are important.
Accountable infrastructure is equally important.
There is something almost absurd about the scale of the AI problem Deezer is now describing. In January 2025, the company said around 10,000 fully AI-generated tracks were being delivered every day. By April, it reported more than 20,000 daily. By January 2026, it was reporting around 60,000. By April, almost 75,000. By June, peak daily deliveries exceeded 90,000.
This is not simply an AI problem. It is the logical endpoint of a digital distribution system in which the marginal cost of producing and delivering another recording has collapsed. The industry built an infrastructure capable of accepting practically unlimited content, then discovered that practically unlimited content could be generated faster than human beings could meaningfully consume it. The platforms helped create the economic environment in which quantity became a competitive weapon. Now they are building algorithms to defend themselves from the consequences of that abundance.
That is not a minor contradiction.
It is the central contradiction of modern streaming.
Nobody should pretend Deezer invented generative AI. It did not. But Deezer participates in the wider platform economy that made digital distribution frictionless, scalable and increasingly detached from physical limitations. When a song can be created, delivered and duplicated with almost no marginal cost, the system naturally attracts people looking for ways to exploit volume. The streaming company then has to build more sophisticated filters to identify the resulting abuse. That creates another technology layer, which produces more data, more classifications and more corporate control. The industry ends up in a strange feedback loop where the solution to platform abuse is often another platform-controlled technical system.
The artist remains caught between both layers.
This is where criticism has to remain intellectually honest. Deezer has uncovered evidence of serious manipulation, and the company’s claim that up to 85% of streams associated with fully AI-generated tracks were fraudulent in 2025 is extremely significant. It would be irresponsible to pretend that streaming fraud is imaginary simply because Deezer is a corporation. Artists genuinely lose when artificial streams distort the royalty economy. The problem is that legitimate fraud prevention can become a convenient justification for expanding the platform’s authority without sufficient transparency. An artist can simultaneously support aggressive action against bots and demand strict limits on corporate discretion. Those positions are not contradictory. They are necessary.
The industry needs stronger fraud detection.
It also needs stronger protection against bad decisions made in the name of fraud detection.
One of the most revealing developments is that Deezer is no longer merely using its AI detection system internally. Since January 2026, it has been commercially licensing the technology to organisations across the music ecosystem, including streaming services, rights organisations and distributors, and says the system has also been used by Billboard. This creates an entirely new commercial incentive. The problem of AI-generated music is no longer only something Deezer needs to manage. Detection itself has become a product. That does not mean the technology is illegitimate.
It does mean artists should understand the economic structure surrounding it. A corporation has developed a tool to solve a problem, demonstrated its usefulness and created a market around that solution. Once detection becomes infrastructure, the standards embedded within that technology can spread far beyond one platform.
Deezer’s definition of the problem can therefore travel with the technology.
This is the question the industry keeps avoiding because it is uncomfortable. Deezer says it detects fully AI-generated music rather than simply policing any use of AI tools. That distinction is welcome. But future production will become increasingly hybrid. A human may write lyrics, perform vocals, play guitar, programme drums, use AI for sound design, edit the result manually and then master the final recording using machine-learning software. Another artist may generate almost everything and make only superficial edits. A third may intentionally create music that mimics the sonic characteristics of generative systems. Where do those recordings sit?
The technology will have to decide.
And whenever technology has to decide, somebody decides what the technology is allowed to mean.
A streaming platform does not need to hate artists to create an environment that is structurally bad for them. It only needs to have more power than they do. Deezer can change policies. Deezer can change detection systems. Deezer can change recommendation rules. Deezer can change how long inactive content remains available. Deezer can alter royalty mechanisms. Deezer can introduce new labels. Deezer can license technology to third parties. An individual musician cannot realistically negotiate those terms. That is the fundamental imbalance. The artist is technically a participant in the platform while remaining economically dependent upon decisions they cannot meaningfully negotiate.
That is not independence.
It is participation under somebody else’s rules.
In July 2026, Deezer announced that it would take down generative-AI tracks used for streaming fraud and AI tracks that had not been streamed for at least six months. From a platform-management perspective, catalog cleanup can sound sensible. From an artist perspective, however, any policy that removes recordings should immediately trigger questions about preservation, notice, appeal and control. Music does not necessarily behave like ordinary disposable content. An obscure track can become important years after release.
An underground recording can develop a following slowly. A release can have cultural value despite low streaming numbers. A platform may reasonably need to manage storage and abuse, but artists should never confuse platform retention policies with permanent ownership of their creative history.
Your archive should not depend on somebody else’s storage policy.
The modern musician should keep original masters, stems where possible, session files, metadata, registration information, distributor records, artwork, release dates and documentation showing how the recording was created. This becomes especially important as automated systems increasingly classify content. If a platform can make an automated claim about your music, the artist should be capable of producing evidence about its creation. This is not because artists should have to prove their humanity to corporations. It is because the balance of power has already shifted toward platforms possessing enormous quantities of data about the content they host. Artists need their own records to avoid becoming completely dependent upon corporate databases.
Documentation is now part of artistic independence.
The AI controversy makes the problem visible, but the underlying issue is much broader. Artists already depend on platforms to deliver metadata correctly, connect releases to the correct artist pages, surface music in search, process streams accurately, identify fraudulent activity and maintain access to their catalogue. Deezer’s own community forum contains examples of artists reporting delays in obtaining artist-page access and problems involving music being associated with the wrong artist identity.
One 2025 case involved an artist waiting weeks for approval while trying to correct an incorrect artist-page association. One anecdote does not prove a systemic failure, but it illustrates the practical reality: the independent artist can encounter administrative systems that are slow, opaque or difficult to navigate, while the platform remains the institution controlling the relevant infrastructure.
That gap matters.
Deezer’s annual reporting makes another structural fact clear: the platform maintains contractual relationships with major recording providers including Sony Music Entertainment, Universal Music Group and Warner Music Group, as well as Merlin representing independent labels and distributors. This does not mean Deezer is secretly controlled by the majors, nor does it prove wrongdoing. It does demonstrate that the streaming ecosystem is not simply a direct relationship between a small artist and a neutral technology company. The platform operates inside a much larger commercial network involving major rights holders, distributors, collecting societies, advertisers, investors and technology providers. When an independent artist uploads a record, they are entering an ecosystem whose negotiating power is overwhelmingly larger than their own.
That reality should destroy the fantasy that streaming platforms are neutral pipes.
They are institutions.
Listeners often imagine that the platform simply provides access to music they choose. In reality, the platform mediates what they discover. Recommendations, playlists, search, labels and algorithmic categorisation influence what gets heard. Deezer says AI-generated tracks are removed from algorithmic and editorial recommendations, meaning the company’s classification system can directly influence what listeners encounter. That may protect consumers from being overwhelmed by synthetic spam, but it also demonstrates the enormous cultural power embedded in the platform. Deezer does not merely store music. It helps construct the environment through which music becomes visible.
And whoever controls visibility controls part of culture.
Every platform can describe itself as artist-friendly. The phrase means almost nothing without measurable standards behind it. How quickly are disputes resolved? How transparent are moderation decisions? What evidence is supplied? How often are mistakes corrected? Can artists appeal? Who reviews appeals? What happens to disputed content during review? How are algorithmic penalties explained? How are royalties audited? What information does an artist receive about suspicious streams? How can an artist challenge a classification? How much of the decision-making process is automated?
Those questions are more meaningful than corporate slogans.
This is the most important criticism to make. Deezer’s messaging is not crude. It is sophisticated. The company speaks about artist rights, transparency, fairness, fraud prevention and protecting creators. It publishes statistics. It conducts research. It develops technical systems. It partners with rights organisations. It talks about artist-centric remuneration. That makes the platform more interesting to criticise, not less. The real question is whether the corporate architecture underneath the language gives artists meaningful power or simply creates a more sophisticated version of platform dependency.
The problem with modern corporate control is that it rarely arrives announcing itself as exploitation.
It arrives announcing itself as protection.
If Deezer labels your music as AI-generated, the artist should be able to ask what that means, what evidence supports the classification and what practical consequences follow. If a stream is considered fraudulent, the artist should understand the basis for the decision and what mechanism exists to contest it. If an artist page is incorrectly associated with somebody else, there should be a clear correction process. If content is removed, there should be transparent notice and meaningful appeal. These are not unreasonable demands. They are basic principles of procedural fairness.
The platform does not need to reveal every detail of its anti-fraud technology to satisfy them.
It does need to demonstrate that the artist has somewhere to go when the machine gets it wrong.
A detector does not lose a release opportunity. An algorithm does not watch a catalogue disappear. A recommendation engine does not spend years building a creative identity. A fraud system does not understand the emotional significance of a recording. The consequences fall on people. That is why human accountability must remain above automated enforcement. The more powerful the algorithm becomes, the more important the human review process becomes. Automation should reduce administrative burden. It should not eliminate responsibility.
Someone must ultimately answer for the decision.
This is the practical lesson. Use streaming platforms. Exploit their reach. Make your music available where listeners expect to find it. But do not confuse distribution with ownership of your audience. Keep your own website. Maintain your mailing list. Preserve your catalogue independently. Keep direct communication with listeners. Sell directly when possible. Build physical products where appropriate. Offer downloads, merchandise and exclusive releases outside the streaming ecosystem. Make sure that if Deezer changes its rules tomorrow, your entire relationship with your audience does not disappear with the change.
The platform should be one route to the audience.
It should never be the only route.
For an independent project like UNIDARK, this distinction is fundamental. Streaming platforms can provide exposure, but direct ownership provides something much more valuable: a relationship that does not depend entirely on somebody else’s algorithm. The Official UNIDARK Store gives listeners a direct way to support the project rather than reducing the entire relationship to another stream inside a corporate platform. The Official UNIDARK Hub provides a central destination outside any single streaming service, allowing the project to maintain its own identity, catalogue and network of links. That is what independence should mean in the digital age. Not pretending platforms do not exist, but refusing to allow them to become indispensable.
This is the contradiction at the centre of Deezer’s current position. The company says it is fighting fraudulent AI music to protect legitimate creators. It says its detection technology is highly accurate. It says fraudulent streams are removed from royalty calculations. It says AI-generated music is excluded from recommendations. It has built a commercial detection product around the technology. All of those measures may have legitimate purposes. But none of them removes the need for accountability. In fact, they make accountability more important because Deezer’s decisions increasingly have consequences beyond simple playback.
The more power the platform acquires, the more scrutiny it deserves.
The old music industry was brutally hierarchical. Artists needed labels, distributors, radio stations and retailers. Digital platforms promised to break those chains. In some respects, they did. An artist can now distribute music globally without signing a traditional record deal. But a different dependency emerged. Instead of asking a label for permission to release, artists increasingly depend on platforms for visibility, discovery, analytics, access and audience attention. The gatekeeper did not disappear.
It became an interface.
When a human executive rejects you, you know you were rejected. When an algorithm quietly deprioritises your music, you may never know. When a playlist editor says no, there is a decision. When an automated system changes your discoverability, the outcome can look like ordinary market performance. This is what makes platform power so difficult to challenge. The artist sees the result but not necessarily the mechanism. The company sees the mechanism but does not necessarily experience the consequence.
That is the accountability gap.
A company genuinely committed to creators should not be frightened by demands for transparency. It should publish clear definitions. It should explain the limits of its detection systems. It should establish meaningful appeals. It should publish information about error correction. It should explain how classifications affect recommendations. It should distinguish fully generated music from legitimate AI-assisted production. It should make it possible for independent artists to resolve disputes without needing corporate representation.
The strongest protection for artists is not a promise.
It is a process.
The deeper rot in streaming is the transformation of music into a system of measurable behavioural data where the creator increasingly becomes one participant inside an enormous technical machine. Songs become assets. Listeners become engagement statistics. Artists become catalogues. Streams become data points. Recommendations become optimisation problems. Fraud becomes an adversarial technical challenge. Even creativity becomes something platforms increasingly attempt to classify automatically.
The danger is not technology itself.
The danger is allowing technology to become more important than the people it was supposedly built to serve.
Deezer’s AI statistics are a warning about the future of every platform. The company reported more than 13.4 million AI-generated tracks detected and tagged in 2025, followed by approximately 90,000 AI-generated deliveries per day at peak levels in June 2026. That volume guarantees further automation. Human beings cannot manually inspect an ocean of content arriving every day. More classification will be automated. More moderation will be automated. More fraud detection will be automated. More recommendations will be automated.
That makes the question of accountability urgent now.
Not after the system becomes impossible to escape.
The independent musician who understands this early has an enormous advantage. Build the catalogue independently. Keep the masters. Keep the documentation. Maintain direct contact with listeners. Own a destination. Create a store. Develop multiple discovery channels. Treat streaming as distribution rather than identity. If a platform disappears, you should still exist. If an algorithm changes, your audience should still be reachable. If a detector makes a mistake, you should have enough independence to survive the dispute.
That is not anti-streaming.
It is pro-artist.
This distinction matters. Deezer is a commercial company. Its responsibility is to operate its business. It is not your manager, your label, your lawyer, your archivist or your artistic representative. The moment artists start expecting platforms to behave like advocates for their careers, they have misunderstood the relationship. Deezer can offer useful infrastructure. It can provide discovery. It can develop tools that combat genuine fraud. But the artist must remain responsible for protecting the larger creative business.
Never outsource the entire future of your music to a company whose first obligation is to its own platform.
If Deezer wants artists to trust its algorithms, then Deezer should be willing to accept scrutiny of those algorithms. If it wants to classify music, it should explain the categories. If it wants to remove fraudulent streams, it should provide meaningful information about disputes. If it wants to label music, it should give artists a credible path to challenge mistakes. If it wants to call itself artist-centric, the artist should be more than the justification used in corporate announcements.
Because “we are protecting artists” is not evidence of protection.
The evidence is what happens when the platform makes a mistake.
This is the uncomfortable lesson Deezer’s rise as an AI gatekeeper makes impossible to ignore. The modern music platform can become distributor, discovery engine, classifier, fraud investigator, royalty administrator and technological authority at the same time. That concentration may be efficient. It may even be necessary in certain areas. But efficiency is not the same thing as accountability.
The artist needs leverage.
The artist needs alternatives.
The artist needs ownership.
And the artist needs a direct relationship with the people who actually care about the music.
That is why building an independent destination matters more than ever. UNIDARK’s Official UNIDARK Store is one example of that principle in practice: give listeners a direct route to the work, rather than making every relationship dependent on a streaming platform. The Official UNIDARK Hub provides another layer of independence by keeping the project’s wider ecosystem in one place.
Use Deezer if it helps you.
Use Spotify if it helps you.
Use every platform that delivers value.
But never mistake access to a platform for ownership of your career.
Deezer has exposed a genuine crisis in streaming. Its own numbers show how quickly synthetic music and fraudulent activity are scaling. Its detection research may become valuable infrastructure for the wider industry. Its artist-centric payment experiments deserve examination. But the more authority Deezer accumulates, the less acceptable it becomes to treat corporate declarations as sufficient evidence of accountability.
The platform should be judged by what happens when its systems are challenged.
It should be judged by how independent artists are treated when something goes wrong.
It should be judged by whether artists can obtain meaningful explanations.
It should be judged by whether the company accepts that an algorithm can be wrong.
And it should be judged by whether “artist protection” means actual power for artists or simply another layer of corporate control presented in friendlier language.
The music industry has spent decades learning what happens when creators surrender too much control to institutions.
The streaming era was supposed to change that.
It should not end with artists discovering that the old gatekeeper has simply been replaced by a machine with a corporate logo.
UNIDARK, also known as Morning Star, is a UK-based independent extreme metal producer and the creator of Blackdeathgrin Metal — an original extreme metal genre combining elements of black metal, death metal, deathcore, and grindcore.
Explore the Official UNIDARK Hub
The central directory for music, releases, streaming platforms, videos, lyrics, Blackdeathgrin Metal information, and all official project links.
→ Official UNIDARK Hub
Explore Official Lyrics & Song Meanings
Discover the complete UNIDARK lyrics archive, song interpretations, and the concepts behind the music.
→ Lyrics Archive
Watch Official UNIDARK Music Videos
Experience the visual side of the project through official releases and video content.
→ Video Archive
Explore Metal Insights
Read more deep articles covering extreme metal genres, underground music, industry topics, production, platforms, culture, and the evolution of heavy music.
→ Metal Insights
Explore Sync Licensing
Read more deep articles covering music licensing, sync rights, music for film, television, games, advertising, trailers, independent production, hybrid music production, catalogue clearance, and the evolving relationship between music and visual media.
→ Sync Licensing
Explore the UNIDARK FAQ Archive
Find answers about UNIDARK, Morning Star, Blackdeathgrin Metal, releases, the creative process, and the philosophy behind the project.
→ UNIDARK FAQ
Support UNIDARK Directly
Support independent extreme metal creation through the official UNIDARK store.
→ Official UNIDARK Store
Join the Official UNIDARK Transmission
Subscribe for new releases, Metal Insights articles, official announcements and updates directly from UNIDARK. Follow the evolution of Blackdeathgrin Metal and the continuing development of the UNIDARK project.
No posts
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.