<\/p>\n
AI music generator Suno is fighting back against copyright infringement claims from independent artists \u2013 with a bold legal argument that could reshape the AI music debate.<\/p>\n
The indie artist class action lawsuit \u2013 filed by Anthony Justice<\/a> and 5th Wheel Records in June 2025 \u2013 specifically targets the impact of AI training on the music of independent artists.<\/p>\n Yet in a motion to dismiss<\/a> filed in federal court on August 18, Suno argues that the indie artist lawsuit \u201cfails as a matter of law\u201d and should be dismissed.<\/p>\n That\u2019s because, it says, music made on Suno doesn\u2019t actually sample existing recordings \u2013 regardless of what music was used to train its AI model.<\/p>\n This represents a potentially game-changing legal strategy: Suno claims that even if its AI learned<\/em> from copyrighted songs, the outputs it generates are entirely new sounds<\/em> that cannot infringe existing recordings under copyright law.<\/p>\n Suno\u2019s filing, obtained by MBW, can be read in full here<\/a>.<\/p>\n The motion represents the latest salvo in ongoing legal battles over AI training on copyrighted music, with Suno already defending against a separate lawsuit filed by major labels<\/a> Universal Music Group<\/a>, Sony Music Entertainment<\/a>, and Warner Music Group<\/a> in June 2024.<\/p>\n That RIAA<\/a>-backed lawsuit remains active and focuses solely on fair use arguments around AI training.<\/p>\n Suno\u2019s<\/strong> motion argues that the independent artist case is essentially a copycat lawsuit that follows the RIAA case \u2013 but adds a problematic new legal theory.<\/p>\n Unlike the ongoing RIAA case, which focuses solely on whether AI training<\/em> practices are legal, Justice and other independent artists are trying to argue that specific songs generated<\/em> by Suno\u2019s AI (\u2018outputs\u2019) substantially copy their original recordings.<\/p>\n At the heart of Suno\u2019s argument is Section 114(b) of the Copyright Act<\/strong>, which has special protection rules for sound recordings that are different from all other types of copyrighted works.<\/p>\n \u201cPlaintiffs would need to allege that the output did not merely sound similar or even identical to the original, but that it actually contained a \u2018sample\u2019 from the original, in the manner of a collage\u2026 that is categorically not how Suno\u2019s tool works. It exclusively generates new sounds, rather than stitching together samples.\u201d<\/p>\n Suno motion to dismiss<\/span><\/p>\n<\/blockquote>\n Suno\u2019s lawyers argue: \u201cWhat [the Copyright Act stipulation] means for this litigation is that in order to plead that a given output of Suno\u2019s AI tool infringed the rights in one of the works-in-suit, Plaintiffs would need to allege that the output did not merely sound similar or even identical to the original, but that it actually contained a \u2018sample\u2019 from the original, in the manner of a collage.\u201d<\/p>\n The filing emphasizes: \u201cBut\u2026 that is categorically not how Suno\u2019s tool works. It exclusively generates new sounds, rather than stitching together samples.\u201d<\/p>\n It reiterates: \u201cThe only way a new recording can infringe the rights in a pre-existing one [under the US Copyright Act] is if it borrows the actual sounds of the original\u2026 that is not how Suno\u2019s tool works at all, as Plaintiffs well know. No Suno output contains anything like a \u2018sample\u2019 from a recording in the training set, so no Suno output can infringe the rights in anything in the training set, as a matter of law.\u201d<\/p>\n This technical distinction forms the core of Suno\u2019s<\/strong> bold legal argument that its AI-generated music cannot legally infringe existing sound recordings, regardless of how similar the outputs might sound to copyrighted works \u2013 and regardless of what music was used to train its AI model.<\/p>\n Significantly, Suno\u2019s filing highlights two recent federal court victories for AI companies on fair use grounds, noting that \u201cnine days after this Complaint was filed, the first federal court to address that overarching issue decided on summary judgment that using millions of copyrighted books to train a generative AI model was fair use.\u201d<\/p>\n The motion references the June 2025 ruling in Bartz v. Anthropic PBC<\/strong>, where a California federal judge found AI training to be protected fair use, followed by a similar decision in Kadrey v. Meta Platforms<\/strong>.<\/p>\n The independent artists\u2019 lawsuit relies heavily on a May 2025 report from the U.S. Copyright Office that questions whether AI training on copyrighted works qualifies as fair use, particularly for music generation.<\/p>\n The Copyright Office report states that \u201cthe fair use doctrine does not excuse unauthorized training on expressive works (e.g., music) particularly when those works are used to generate substitutional outputs that may replace the originals in the relevant marketplace.\u201d<\/p>\n However, Suno\u2019s motion to dismiss notes that subsequent court rulings have sided with AI companies on fair use grounds, potentially undermining the Copyright Office\u2019s analysis.<\/p>\n In August 2024, both Suno and Udio argued<\/a> that their use of copyrighted materials fell under fair use protections while acknowledging they trained their models on copyrighted recordings.<\/p>\n The broader legal battle over AI music generation continues to unfold across multiple federal courts, with the outcome likely to shape how the music industry approaches licensing and partnership deals with generative AI platforms going forward.<\/p>\n Suno is also facing a copyright lawsuit<\/a> from Germany\u2019s GEMA<\/a> collection society, filed in January 2025, adding to its international legal challenges.<\/p>\nSuno argues indies\u2019 lawsuit introduces legally flawed claims<\/h6>\n
\n
U.S. Copyright Office report cited in original lawsuit<\/h6>\n