AI Music Digest

A new Morgan Stanley survey on American audio habits has found that 60% of listeners aged 18-29 and 55% of those aged 30-44 now regularly consume AI-generated music, averaging 2.5-3 hours per week.

The survey marks the first time the financial firm included questions about AI music in its annual audio research. Analyst Benjamin Swinburne noted: “What we found surprised us.” YouTube and TikTok emerged as the primary platforms for AI music consumption—not traditional streaming services.

This finding contrasts sharply with data from Deezer, which reported that while AI tracks comprise 34% of new uploads, they account for only 0.5% of actual streams on its platform. The discrepancy suggests AI music is gaining significant traction on social media while traditional streaming remains dominated by human artists.

Why It Matters: This data reveals a generational and platform divide in AI music adoption. If younger listeners are consuming hours of AI music weekly on social platforms, traditional streaming metrics may be understating AI’s actual cultural footprint.

Universal Music Group has announced a strategic partnership with NVIDIA, marking the first collaboration between the world’s largest record label and the leading AI chipmaker. The companies will develop what they call “responsible AI” tools for music discovery, creation, and fan engagement.

Central to the partnership is NVIDIA’s Music Flamingo model, which can analyze full-length songs up to 15 minutes, identifying chord progressions, instruments, lyrics, and cultural context. The model was trained on approximately 2 million songs across over 100 genres.

UMG and NVIDIA will establish an artist incubator at Abbey Road Studios and Capitol Studios, bringing together artists, songwriters, and producers to co-design AI tools. The incubator is explicitly positioned as “a direct antidote to generic, ‘AI slop’ outputs.”

UMG Chairman Sir Lucian Grainge emphasized the company’s boundaries: “We will NOT license any model that uses an artist’s voice or generates new songs without their consent.”

Why It Matters: By partnering with the world’s most valuable company ($4.56 trillion market cap), UMG is signaling that major labels see AI as inevitable—but want to shape it on their terms, with artist consent and compensation built in.

Napster has shut down its music streaming service after 25 years, pivoting entirely to an AI-focused platform. Subscribers received notifications stating: “Napster is no longer a music streaming service. We’ve become an AI platform for creating and experiencing music in new ways.”

The company, acquired by Infinite Reality in 2025 for $207 million, instructed users to export their playlists to competing services before the transition. Its new focus is “AI Companions”—tools that understand users’ music taste and help them discover and create content.

Napster’s CTO framed the shift as history repeating: “AI is basically making all of us creators. We’re not just consumers of content—we can all create content at a higher fidelity.” He argued Napster is once again identifying consumer demands the industry wasn’t addressing.

Why It Matters: The company that disrupted the music industry in 1999 is betting its future on AI disrupting it again. Whether this pivot succeeds or fails, it signals that some industry players see AI creation as the future of music consumption.

Darkwave artist Anders Manga has filed a copyright lawsuit against Stability AI and licensing intermediary AudioSparx, alleging his music was used to train the Stable Audio model despite explicit opt-out requests.

The complaint, filed December 29 in North Carolina federal court, claims Manga’s works were covered under a 2015 licensing agreement that predated AI and did not authorize training use. When he requested removal, AudioSparx allegedly denied his requests, and Stability AI deferred to its partner.

Stability AI maintains that Stable Audio was “built specifically for professionals and trained exclusively on licensed data.” However, the case raises fundamental questions about opt-out mechanisms—often touted by AI companies as sufficient creator protection.

Why It Matters: This case could test whether “we licensed it from a partner” is a valid defense when the partner’s authority is contested. If opt-out mechanisms prove unenforceable, AI companies may need to shift toward explicit, affirmative consent for training.