{"id":498,"date":"2026-09-26T08:08:13","date_gmt":"2026-09-26T00:08:13","guid":{"rendered":"https:\/\/aidashxp.com\/sony-umg-sue-suno-model-laundering\/"},"modified":"2026-09-26T08:08:13","modified_gmt":"2026-09-26T00:08:13","slug":"sony-umg-sue-suno-model-laundering","status":"publish","type":"post","link":"https:\/\/aidashxp.com\/en\/sony-umg-sue-suno-model-laundering\/","title":{"rendered":"Sony and Universal Music Sue Suno Again: Alleging Model Money Laundering, AI Music Copyright War Escalates"},"content":{"rendered":"

From \u201cVoice Assistant\u201d to \u201cVoice Agent\u201d: An UpgradeOpenAI Agent Out of Control: 53 User Photos Leaked to the Public Internet, Raising New Alarms About Agent Security<\/strong>September 26, 2026 Suno<\/a><\/strong> The copyright battle in AI music has reignited. On September 25, according to The Verge,Sony Music (Sony) and Universal Music Group (UMG)<\/strong>have once again filed lawsuits against AI music generation companies.<\/p>\n\n\n\n

This round\u2019s central allegation centers on a new term\u2014<\/h2>\n\n\n\n

\u201cmodel laundering\u201dSuno<\/a> This marks the second legal offensive by major record labels against AI music, following their first courtroom confrontation in 2024.What is \u201cmodel laundering\u201d?<\/strong>\u201cModel laundering\u201d is a sharp concept coined by record labels, targeting the core controversy in AI music:<\/p>\n\n\n\n

Companies like these train models on massive volumes of copyrighted musical works, then generate music that \u201cappears entirely new.\u201d Plaintiffs argue this effectively \u201claunder\u201d copyrighted content into generated outputs that cannot be traced back, thereby circumventing copyright law.<\/p>\n\n\n\n

For example: if an AI-generated song bears strong similarities to a particular artist\u2019s work in melody or arrangement\u2014but falls short of verbatim copying\u2014the original creator often struggles to hold the AI company liable under traditional copyright law. The term \u201cmodel laundering\u201d aims precisely to dismantle this gray zone where something is \u201ctechnically non-infringing yet substantively copied.\u201dThe conceptual power of \u201cmodel laundering\u201d lies in its refusal to dwell on whether \u201cthe AI copied verbatim.\u201d Instead, it directly targets<\/strong>the training data itself:<\/p>\n\n\n\n

if your model was trained on unauthorized works, then no matter how \u201coriginal\u201d the output appears, its origin is already \u201ctainted.\u201d<\/h2>\n\n\n\n

Why a second lawsuit? Suno<\/a>From \u201cVoice Assistant\u201d to \u201cVoice Agent\u201d: An Upgrade Suno<\/a> OpenAI Agent Out of Control: 53 User Photos Leaked to the Public Web, Raising New Alarm Bells for Agent Security<\/p>\n\n\n\n

September 26, 2026<\/p>\n\n\n\n

, accusing it of training its models on vast volumes of copyrighted music without authorization. At the time,<\/h2>\n\n\n\n

its defense was \u201cAI learning is like humans listening to songs to learn music\u2014it qualifies as fair use.\u201d Suno<\/a> Today, major record labels have struck again, indicating that the first lawsuit failed to resolve the issue at its root\u2014and underscoring that the explosive rise of AI music has already tangibly disrupted the recording industry\u2019s interests. This time, they\u2019re framing the case around \u201cmodel laundering,\u201d aiming to pry open a new legal avenue\u2014not debating \u201chow similar\u201d outputs are, but directly challenging \u201cthe legality of training data.\u201dThe Copyright Dilemma of AI Music<\/a>This is not just<\/p>\n\n\n\n

a lawsuit against one company, but a shared challenge confronting the entire generative AI field. As previously reported by this site,a Microsoft executive labeled AI web scraping \u201cunprecedented theft,\u201d<\/strong>and the New York Times copyright case has thrust the question of \u201cwhether training data infringes copyright\u201d into the spotlight.<\/p>\n\n\n\n

AI music, AI art generation, and AI writing are fundamentally the same issue: Suno<\/a> Should model training data be licensed\u2014and compensated\u2014for use by original creators?<\/p>\n\n\n\n

What does it mean for users and the industry?<\/h2>\n\n\n\n