LANDR<\/b> says the volume of artificial streams flagged by Spotify<\/b> across the music it distributes fell by 95%<\/b> between Q1 2025<\/b> and Q2 2026<\/b>.<\/p>\n
That’s according to a press release issued by the Montreal-based distribution and music creation platform on Tuesday (September 22<\/b>), in which the firm announced an expansion of a streaming fraud detection system it built with Quebec AI lab Mila<\/b>.<\/p>\n
Mila<\/b> is a non-profit AI research institute in Montreal<\/b>, founded by Yoshua Bengio<\/b>, who shared the 2018<\/b> A.M. Turing Award<\/a> for his work on deep learning.<\/p>\n The initial machine-learning models were developed and deployed by \u00c9douard Demotes-Mainard<\/b> as part of the collaboration between Mila \u2013 Quebec AI Institute<\/b> and LANDR<\/b><\/span><\/a>. LANDR<\/b>‘s anti-fraud team has since operated and continued developing the system.<\/p>\n LANDR<\/b> attributes the 95%<\/b> decline in flagged streams to the rollout of multiple fraud-prevention initiatives, rather than to the Mila<\/b> system alone.<\/p>\n One of the models assesses accounts using profile and behavioral data, while the other examines listening activity for patterns inconsistent with a genuine audience, says LANDR<\/b>.<\/p>\n LANDR<\/b> says it has since expanded the system to identify additional risk signals earlier in the distribution process.<\/p>\n Those signals include unusually high release frequency, catalogs that appear overnight, clusters of accounts sharing devices or payout details, and metadata designed to impersonate established artists, according to LANDR<\/b>.<\/p>\n Flagged activity is reviewed by LANDR<\/b>‘s anti-fraud team before any action is taken, the company says.<\/p>\n Accounts confirmed to be fraudulent can be closed and blocked from returning, while confirmed cases help LANDR<\/b> continue improving its detection models, the company added.<\/p>\n “Finding fraud by hand at our scale was a needle-in-a-haystack problem. The first generation of the model turned that into something we could actually manage,” said Tommy Desilets<\/b>, CTO of LANDR<\/b>.<\/p>\n “Since then, we’ve continued improving the system as we identify new patterns of fraudulent activity. Now we’re moving those capabilities earlier in the distribution process, so our team has better signals before suspicious releases reach streaming services.”<\/p>\n “An artificial stream isn’t just somebody gaming a platform. It competes for money from the same royalty pool that pays musicians who are trying to make a living.”<\/p>\n Pascal Pilon<\/b>, LANDR<\/b><\/span><\/p>\n<\/blockquote>\n LANDR<\/b> defines fake streams as plays generated without genuine listener intent, often via bots, stream farms, or compromised accounts.<\/p>\n Fraudulent plays carry a cost for distributors as well as for rightsholders: Spotify<\/b> has charged labels and distributors a per-track fee when it detects “flagrant artificial streaming” on their content since April 2024<\/b>, as previously reported by MBW<\/i><\/span><\/a>.<\/p>\n “An artificial stream isn’t just somebody gaming a platform. It competes for money from the same royalty pool that pays musicians who are trying to make a living,” said Pascal Pilon<\/b>, co-founder and CEO of LANDR<\/b>.<\/p>\n LANDR<\/b> says human review sits between detection and enforcement because an unusual pattern in the data is not proof of fraud.<\/p>\n “We want AI helping our people make better decisions at scale, not replacing judgment,” Pilon<\/b> said. “Sometimes a song suddenly taking off is exactly that: an artist having a really good day. We’re rooting for them, not trying to put a damper on their success.”<\/p>\n Machine learning is one layer of the company’s fraud-prevention system, according to LANDR<\/b>, which also uses audio-recognition technology from ACRCloud<\/b> and Ircam Amplify<\/b>.<\/p>\n Ircam Amplify<\/b> is the commercial arm of IRCAM-Centre Pompidou<\/b>, the Paris acoustic and computer-music research institute founded by composer Pierre Boulez<\/b> and operating since 1977. It launched a detector for AI-generated music in 2024<\/span><\/a>.<\/p>\n ACRCloud<\/b> fingerprints recordings against commercial catalogs, while the Ircam Amplify<\/b> technology checks for altered recordings and for acoustic markers associated with synthetic generation, the company says.<\/p>\n LANDR<\/b> added that it is rolling out identity verification for accounts with frequent release activity.<\/p>\n Artists using the platform declare AI-generated content when they submit a release, with that information carried in the metadata delivered to streaming services \u2013 a disclosure route Spotify<\/b> began building out with DDEX in September 2025<\/span><\/a>.<\/p>\n “Finding fraud by hand at our scale was a needle-in-a-haystack problem. The first generation of the model turned that into something we could actually manage.”<\/p>\n Tommy Desilets<\/b>, LANDR<\/b><\/span><\/p>\n<\/blockquote>\n LANDR<\/b> says its expansion advances the same priorities set out in IFPI<\/b>‘s Streaming Integrity Initiative<\/b>, the anti-fraud standard for distributors that the trade body published on September 14<\/span><\/a>, which calls for safeguards across five areas: Verify<\/b>, Vet<\/b>, Act<\/b>, Share<\/b>, and Measure<\/b>.<\/p>\n MBW<\/i> understands the company has been in discussions about formally joining the initiative at a later date, though no timeline has been set.<\/p>\n LANDR<\/b> says its platform serves more than 8 million<\/b> musicians, with services spanning distribution, promotion, monetization, mastering, samples, plugins, and collaboration.<\/p>\n\n
\n