Grammy-winning singer SZA has publicly condemned AI music company Suno after discovering that 238 of her songs, including unreleased tracks, were used to train its generative music models. The artist called the practice "disgusting" and warned that AI music tools pose what she described as a "dangerous new threat" to Black musicians in particular.
The outburst, which spread rapidly across major music outlets including Variety, VICE, NME, Complex, and Stereogum over the weekend of June 20-21, 2026, marks one of the highest-profile artist revolts yet against generative AI training practices in the music industry. For more context on this story, see our ongoing AI industry coverage.
238 Songs, Including Unreleased Material
According to reporting by TheGrio, Yahoo, and NME, SZA learned that 238 of her recordings had been ingested by AI systems built to generate music. Variety and the news aggregator inkl confirmed that the figure included unreleased tracks, raising the stakes beyond the familiar debate over catalog licensing into questions about how leaked or unpublished material enters training datasets.
VICE identified the company at the center of the dispute as Suno, one of the leading generative AI music startups. The outlet reported SZA's blunt verdict: there was "nothing you could ever say to me to make this okay."
The scale of the figure, 238 songs, surprised many observers because it suggests broad ingestion of an artist's catalog rather than isolated tracks. For a major-label artist whose official discography spans several albums, the number implies that the bulk of her released work, plus material that never reached streaming services, was swept into the model.
'Disgusting' and 'Degenerate'
SZA did not limit her criticism to the technology itself. Complex reported that she labeled musicians who support AI-generated music with a profanity, calling the trend "disgusting" and "degenerate." Variety and NME echoed the characterization, with NME reporting that she hit out at "disgusting" AI music after learning of the training data.
The frustration reflects a growing fault line inside the music industry, where some creators and producers have embraced AI tools for ideation and production while others argue that models trained on copyrighted recordings without consent amount to systematic appropriation.
Calling Out Diplo and Suno's Investors
SZA expanded her criticism beyond Suno's technology to individual artists with financial ties to the company. Stereogum reported that she "warns AI is exploiting Black artists" and named producer Diplo specifically. The UK's International Business Times carried the same account under the headline that SZA "slams Diplo's equity in Suno AI" and says Black music "now faces a dangerous new threat."
The invocation of Diplo is significant because it shifts the conversation from an abstract debate about algorithms to concrete questions about who profits when a model is trained on other artists' work. If prominent producers hold equity in the very companies ingesting peers' catalogs, SZA's argument goes, the financial incentives run against the artists whose recordings built the models.
SZA's framing, that AI music disproportionately threatens Black artists, taps into a broader concern raised by musicians and advocates across R&B, hip-hop, and soul, genres that are heavily sampled and now heavily trained on. Critics argue that because these genres are among the most data-rich and commercially influential, their creators bear a disproportionate share of the risk when generative models commodify their styles.
A Industry-Wide Reckoning Over Training Data
The clash is the latest flashpoint in a legal and ethical battle that has engulfed the generative music sector. Suno and its competitors have faced lawsuits from major record labels alleging that their models were trained on copyrighted recordings without licenses. The companies have generally argued that their use of training data is permissible, though courts and regulators are still drawing the lines.
What makes SZA's case resonate is the specificity: a named artist, a named company, a precise song count, and the revelation that unreleased material was involved. Those details give the dispute a human face that abstract legal filings often lack, and they are likely to fuel further pressure on AI music firms to disclose and license their training data.
What Comes Next
The episode is unlikely to be the last. As generative music tools grow more capable and widely used, tensions between model builders and recording artists are expected to intensify. Several outcomes are possible: clearer disclosure of training datasets, voluntary licensing deals with major labels, fresh litigation, or new legislation clarifying when copyrighted recordings can be used to train AI.
For now, SZA's stance is unambiguous. As reported across multiple outlets, she views the training of models on her catalog, especially unreleased songs, as a line that should not have been crossed. Her comments add a prominent, mainstream voice to a cause that has until now been championed mostly by industry trade groups and a coalition of independent artists.
Whether that pressure forces changes at Suno and its peers remains to be seen. But the message from one of music's biggest stars is clear: artists are paying attention to what feeds the machines, and they are not happy with what they are finding.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
