Behind the tech: Leveraging ACR for AI training data

WRITTEN BY Jaclyn Petrovich
Apr 30, 2024

The rise of AI-generated technologies has sparked many copyright questions that have yet to be answered. Currently in the U.S., AI-generated content isn’t copyrightable, models potentially are trained on copyrighted content without licenses, and it’s unknown whether an artist could sue for name and likeness if their voice is used in AI content. The lines are not only blurry, but being challenged more and more every day. 

When it comes to copyrighted content, especially music, AI companies need ways to safeguard their technology from potential legal troubles. With Pex, AI platforms can validate that their training data does not contain unlicensed copyrighted works and ensure that the output does not match existing recordings and compositions.

How Pex technology can detect copyrighted works for generative AI companies:

  • Compare your training data against a registry of copyrighted music where ACR technology can then identify any matching audios or melodies
  • With phonetic matching, new recordings of lyrics can be traced back to their original writing, so that use of copyrighted lyrics can be identified in outputs
  • Singer identification and matching can determine if two recordings have the same voice. Recording artists, if registered in the reference database, can be identified so AI-generated voice impersonations in outputs can be detected.

Because it will soon not be possible to distinguish AI-generated works from those that are human created, and because the lines between these will be increasingly blurred as musicians continue to use AI to assist the songwriting process, the best way forward in protecting copyright is through content recognition technology.

Download our free eGuide: Real or fake: Identifying AI music and voices to learn more. 

Identify copyrighted works with Pex

Automated Content Recognition (ACR) and Music Recognition Technology (MRT) can be used to identify uses of copyrighted music in AI training data or AI-generated songs. Pex is at the forefront of identifying digital and AI-generated content. We are always improving our technology and finding new ways to identify copyrighted content.  Want to see our tech in action? Reach out to schedule a demo.

Recent stories

New: Check any file to identify music, speech, or silent segments

We help rightsholders solve some pretty technical problems, like finding all versions of a song online even if it’s been sped up, chopped up, or mashed up. We get into the hairy details of cover versions and publisher splits. But, sometimes the problem plaguing...

Vobile Completes Acquisition of Pex

SANTA CLARA, Calif., April 14, 2025  -- Vobile, a global leader in digital content protection and transaction services, today announced it has completed its acquisition of Pex, a leading technology provider of audio content identification. The acquisition enhances...

Music proves problematic for top beauty brands on social media

Beauty may be in the eye of the beholder, but music licensing isn’t subjective. When it comes to adding music to social media posts as a brand, every upload must be licensed for commercial use, or the brand could be liable for copyright infringement. We’ve been...

From trending to compensated: Solving TikTok’s Sound ID problem

Greetings from the transitional space between TikTok existing and TikTok being banned in the US. Right now, TikTok is available, which means we’re analyzing Sound IDs on the platform to help rightsholders get paid. If you missed our last Sound ID piece where we...