Resources
BIPA and your voice recordings: what owners should know
Your voice can count as biometric data. Illinois BIPA is why. If you record or license audio of identifiable people, consent at the source protects you as much as the buyer.
Published 2026-07-22 · 6 min read
Key takeaways
- A voiceprint can count as biometric information under Illinois BIPA, which has governed biometric data since 2008.
- BIPA requires notice and written consent before collection and gives individuals a private right of action.
- Recent voiceprint suits turn on the same point: public availability of a recording is not consent to train on the speaker’s voice.
- Owners who license audio of identifiable people are part of the chain, so missing consent makes them the weak link.
- Consent captured at the source, and kept with the asset, protects the owner and the buyer at once.
Most owners think about copyright when they license audio. Biometric law is the part they miss. A recorded voice can be more than a performance. Where it identifies a specific person, it can count as biometric information under state law, and that brings a different rulebook.
Illinois has the strictest version. If you record, hold, or license audio of identifiable people, BIPA is worth understanding before a buyer ever asks. Consent captured at the source is what protects both sides.
What BIPA covers
The Illinois Biometric Information Privacy Act, 740 ILCS 14, has governed biometric data since 2008. It covers biometric identifiers and biometric information. A voiceprint, the set of measurable characteristics that identify a person from their voice, can fall within it.
BIPA requires notice and written consent before a private entity collects that data. It requires a retention and destruction schedule. And it gives individuals a private right of action, which means a person can sue directly. That last feature is why BIPA drives so much litigation.
Why AI voice work drew the suits
Voice models need voices. Some builders took them from public recordings without consent. In 2026 a coordinated set of class actions was filed by professional voice artists against several large technology companies, alleging their voiceprints were extracted from publicly available audio and used to train commercial voice models without permission. Separate BIPA cases have targeted voice-assistant and transcription features.
The through-line is consent. Public availability is not consent. A recording being reachable online does not mean the speaker agreed to have their voiceprint captured and used to train a model. BIPA treats that gap as actionable.
Why this matters to owners, not just buyers
If you license audio of identifiable people, you are part of the chain. A buyer who trains on it inherits exposure, and they will look back up the chain to whoever supplied it. An owner who cannot show consent becomes the weak link.
This is not a reason to avoid licensing your recordings. It is a reason to do it with consent attached. The owners who capture consent cleanly are the ones buyers can actually work with.
Consent at the source protects both sides
The fix is consent gathered when the recording is made or before it is licensed, not reconstructed afterward. For identifiable speakers, that means specific, documented permission for AI-training use, kept as a record that travels with the asset.
That is how fiund handles it. Every asset carries a signed AI-training licence. Where a person is identifiable, voice and likeness consent is captured separately. The owner keeps ownership and approves the buyer. When consent lives with the asset, both the owner and the buyer can show it on request.
What owners should do now
Know which of your recordings feature identifiable people. Those are the ones biometric law reaches. Keep any release or consent you already have, and note where it is missing. Do not license audio of identifiable speakers without a consent path.
When you record new material, gather consent for AI-training use at the same time. It is far easier to capture at the source than to chase later. Consent is the asset that makes the recording licensable.
Sources
Frequently asked questions
Is a voiceprint really biometric data?
It can be. A voiceprint is the set of measurable characteristics that identify a person from their voice. Where a recording identifies a specific person, that can fall within biometric law such as Illinois BIPA, which carries notice, consent, and retention duties.
I only license recordings I own. Do I still need consent from speakers?
Owning the recording is a copyright matter. Consent from an identifiable speaker is a separate, biometric and likeness matter. If a person is identifiable, you generally need their consent for AI-training use as well as your own rights in the recording.
Does consent at the source really protect me?
It is the strongest protection available. When consent is captured before licensing and kept with the asset, both you and the buyer can produce it if a claim arrives. Reconstructing consent after the fact is far weaker.
Related resources
Want data that clears this in diligence?
Whether you're building a model or sitting on an archive, the first conversation is short and specific.
Send a brief