aipopdb

19 Oct 2026  ·  3,033 works · 708 news indexed

News / Grammy-winning engineers launch UAI, an AI music detector that issues signed certificates

Grammy-winning engineers launch UAI, an AI music detector that issues signed certificates

AIPopDB Newsroom · via musicbusinessworldwide.com · · 2 min read

Read the original ↗
Image for Grammy-winning engineers launch UAI, an AI music detector that issues signed certificates

Image: musicbusinessworldwide.com

RTM Audio, a new company founded in Los Angeles, has released UAI, a detector for AI-generated music built around one idea: a verdict should be evidence, not a probability. Each track is tested twice, once on the production and once on the vocal, and is only marked as AI when both checks agree. Uncertain cases go to a human reviewer instead of being forced into a yes or no, which the founders present as protection for artists wrongly suspected of using AI.

Every result comes with a cryptographically signed certificate tied to the recording's ISRC and checkable against a public key. The company says the certificates are formatted for EU AI Act disclosures and DDEX metadata, that four patent applications are being prepared, and that no other vendor it knows of offers anything similar. Its internal test on 8,236 masters produced five hard flags, a false-positive rate of 0.06%, a figure it promises to put in writing for each client and recalibrate per catalogue.

The founders come from the studio side. Chief technology officer Ohad Nissim is a Grammy-nominated mastering engineer, chief executive Teezio has won two Grammys, operations chief Calin Enache founded audio software firm Mixwave, and music attorney Matt Buser completes the group. Their pitch is that a human artist should never be accused on one system's say-so, and that a score will not survive a courtroom where a signed record might.

The launch leans on a 2025 academic study that tested detectors on 30,000 tracks, pairing human songs with Suno and Udio output. The researchers built their own detectors and compared them with commercial and open-source ones. One commercial tool wrongly flagged 4.7% of human tracks, and its answer could be flipped in seconds just by resampling the file to a lower rate.

The stakes have grown. Deezer reported about 75,000 fully AI tracks a day in April, Believe and TuneCore now block uploads from unlicensed AI services, the NO FAKES Act cleared a Senate committee on June 18, and EU AI Act transparency rules apply from August 2. Labels are building their own tooling too: Warner agreed to buy attribution firm Sureel in June and Sony led a funding round for identity-protection company Vermillio. Deezer, which has tagged AI tracks since 2025, now licenses its detector to partners such as Sacem and released a free public version on June 11. RTM wants to be a neutral layer outside any one label or platform.