Phoenix 24
Her protest intensifies demands for consent, credit and compensation.
LOS ANGELES, United States | June 2026
SZA has launched a forceful attack on artificial intelligence in music after discovering that 238 tracks associated with her name appeared in datasets available for training generative systems. The Grammy-winning singer shared the result of a search using a recently released database designed to help musicians identify whether their work had been collected for AI development. She said some of the material may never have been officially released. Her reaction placed consent and creative ownership at the center of an increasingly confrontational industry debate.
The database examined by SZA brings together information from four music collections accessible to AI developers and contains references to more than 21 million songs. It is not a complete record of everything used by every commercial model, and inclusion does not by itself prove that a particular company trained on every listed track. However, it reveals the enormous quantity of copyrighted music assembled for machine-learning research and product development. For artists, that scale raises urgent questions about how their recordings were obtained and whether permission was requested.
SZA responded with unusually direct language, condemning musicians who support systems built from creative work collected without approval. Her message reflected more than concern about technology replacing performers. She objected to artists becoming involuntary suppliers of the cultural material that allows commercial platforms to generate new songs. In her view, innovation cannot be separated from the conditions under which the underlying music was acquired.
She later focused her criticism on Suno, one of the most prominent platforms capable of generating complete songs from written instructions. SZA also alleged that producer Diplo holds a financial interest in the company and is helping connect it with influential Black songwriters and producers. That claim was presented by the singer and has not been independently established in the reporting surrounding her comments. Her wider argument was that Black creativity is particularly vulnerable to extraction without adequate legal or economic protection.
SZA emphasized the global influence of Black music despite Black Americans representing a minority of the United States population. Genres developed through Black communities have repeatedly shaped international popular culture, from blues, jazz and rock to hip-hop, R&B and contemporary electronic production. The singer warned creators not to surrender that knowledge to systems that may later compete with them. Her criticism connected AI training with a much longer history of cultural appropriation and unequal compensation.
This is not the first time she has challenged artificial intelligence. Earlier in 2026, SZA said she felt she was in a struggle against technology because synthetic versions of emerging Black artists were already circulating before those musicians had fully benefited from their own work. She also criticized AI-generated R&B for producing narrow representations centered on pain and stereotypes. For her, the problem involves both economic extraction and the reduction of complex artistic traditions into predictable patterns.
Suno maintains that its purpose is to expand human creativity rather than reproduce existing performers. Company product chief Jack Brody recently said the platform has built safeguards intended to stop users from uploading or distributing material they do not own. He also said Suno works with audio-identification companies to detect potential misuse. The company describes its approach as original creation by design.
Brody said Suno does not use artists’ names as training metadata and does not want its systems to generate direct replicas of identifiable musicians. The company argues that models should create new compositions rather than recover specific recordings from their training material. Those safeguards address outputs, but they do not completely resolve the separate question of whether copyrighted songs were lawfully collected in the first place. That distinction remains central to the conflict.
Major record companies sued Suno and rival platform Udio in 2024, alleging that copyrighted sound recordings had been copied without authorization to train their models. The companies have disputed the characterization of AI training as straightforward theft and have defended generative technology as a new form of creative development. The cases have become important tests of how existing copyright law applies when machines analyze enormous music catalogs. Their outcomes could reshape licensing across the industry.
Artists and rights organizations argue that training should require permission, attribution and payment. They contend that an AI company should not build a commercial product from recordings whose owners never agreed to participate. Developers respond that machine learning involves analyzing patterns rather than storing a conventional library of songs for public reproduction. Courts must determine where learning ends and legally significant copying begins.
The dispute is especially complicated because a dataset entry may point to a song hosted on services such as YouTube or Spotify without containing the original audio file itself. Automated tools can collect links, metadata or recordings through different technical methods. Some of those methods may violate platform terms even before copyright questions are considered. Transparency about the full training process remains limited.
SZA’s suggestion that unreleased songs appeared in the search results adds another level of concern. Unpublished material may circulate through leaks, private sharing networks or incorrectly labeled files. Its presence would demonstrate that public availability cannot always be assumed. It also raises questions about whether artists can protect works before their commercial release.
Generative music platforms have attracted millions of users because they make composition accessible to people without formal training or expensive equipment. Supporters see them as tools for experimentation, rapid demonstrations and independent production. Critics fear that the same efficiency will flood streaming services, advertising and film licensing markets with low-cost synthetic content. The economic consequences could reach session musicians, composers, singers and producers whose work is already financially precarious.
The controversy is therefore not a simple conflict between artists and technology. Many musicians use software, sampling, digital instruments and machine-assisted editing as part of their practice. The decisive issue is whether the creator controls the process and benefits from it. SZA’s protest rejects a system in which participation is assumed after the work has already been collected.
Her intervention may encourage more artists to search the available datasets and publicly challenge what they find. A collective response could increase pressure for licensing agreements, stronger disclosure rules and the ability to remove works from future training. It could also divide the music community between those willing to collaborate with AI companies and those demanding strict limits. That division is likely to intensify as synthetic songs become more difficult to distinguish from human recordings.
SZA’s language was confrontational, but the underlying demand was precise: creators should decide whether their work helps train commercial systems. Without consent, credit and compensation, technological progress risks repeating familiar patterns of exploitation. The debate will ultimately determine who controls the value produced when human culture becomes machine-readable.
Innovation loses legitimacy when creation is taken without consent. / La innovación pierde legitimidad cuando la creación se toma sin consentimiento.