Spotify removes millions of AI-created songs in effort to combat spam
An internal team at the company is now actively targeting AI-generated audio clutter
By Alfonso Maruccia
July 17, 2026, 13:01
Add TechSpot
Serving tech enthusiasts for over 25 years.
TechSpot means tech analysis and advice you can trust.
SpotiSlop: Advanced language models and sophisticated audio-generation tools are reshaping the conventional music landscape. Digital streaming services like Spotify have become prime targets for AI-driven spammers, though industry bodies are pushing back with new labeling initiatives.
Sam Duboff of Spotify recently disclosed that the platform had to take down 75 million AI-produced tracks during 2025 alone. In his role as Spotify’s senior director and global head of marketing, policy, and music business, Duboff contends that AI has not introduced completely novel methods for spam operations. Still, he notes that generative AI and other machine learning technologies have elevated audio spam to a new level.
Spotify now employs systems aimed at fighting both AI-generated “slop” uploads and data scraping by firms looking for fresh content to feed their generative AI models. Duboff confirmed that the company maintains a “big team” tasked with identifying possible new attack routes that could simplify matters for spammers (or scraping bots).
The executive also shared several striking statistics that underscore the rising frequency of AI-produced music on digital platforms. Each day, bots upload roughly 100,000 distinct “songs” to Spotify’s infrastructure, and a substantial share of these tracks are likely “created” using generative AI services or custom LLM-based configurations.
Duboff acknowledges that not every generative AI track can be automatically written off as slop. The 75 million songs removed last year consisted of low-effort content involving very little human creativity beyond a lazy chatbot prompt. However, “real” artists are increasingly incorporating AI technology into their production processes, which blurs the line between fake music and human-curated work.
Spotify is certainly adopting AI across its operations these days. Earlier this year, co-CEO Gustav Söderström stated that the company’s engineers are essentially letting AI agents handle much of their coding tasks. Spotify also indicated that AI tools approved by music labels are now permissible for creating remixes and covers that can be sold on the platform.
// Related Stories
-
Spotify CEO defends AI music, promises musicians will be adequately compensated -
The first CD recorder was shockingly expensive – guess how much
Beyond Spotify’s difficulties with spammy tracks, generative AI is compelling the entire music sector to devise fresh approaches to the increasingly pressing AI slop problem. The International Federation of the Phonographic Industry, the RIAA, and other major trade organizations recently unveiled a “voluntary” initiative to properly label AI-generated music, providing listeners with a clear signal when a song has been produced entirely via a chatbot prompt. These labels are also meant to identify “AI-assisted” music that was largely created by human artists.


