Biometric consent for speech data

US biometric privacy statutes treat voiceprints as biometric identifiers. The Illinois Biometric Information Privacy Act (BIPA) requires informed written consent before collection and gives individuals a private right of action with statutory damages.

Why it matters to a buyer

"It was publicly available" is not a defense. In May 2026, class actions by broadcast journalists and voice actors alleged that companies extracted voiceprints from public audio to train commercial voice and TTS models without consent. If a speech corpus lacks written consent from the speakers, the exposure is per-person and statutory.

Why it matters to a data owner

Written consent from your speakers is a licensable asset. Suppliers who hold it can serve buyers that BIPA-exposed datasets cannot.

Current legal status

BIPA defines voiceprints as biometric identifiers and requires informed written consent before collection, enforceable through a private right of action with statutory damages. In May 2026, nine class actions filed in federal court in Chicago by broadcast journalists, podcasters, and voice actors alleged that Adobe, Amazon, Apple, ElevenLabs, Google, Meta, Microsoft, NVIDIA, and Samsung extracted voiceprints from publicly available audio to train AI voice models without consent; the suits are in their early stages. Texas (CUBI) and Washington have analogous statutes with different enforcement. Public availability of audio is not consent under these laws.

What fiund does about it

fiund pairs the licence with written, documented consent where speakers are identifiable — the artifact these statutes require and these suits allege was missing.

Sources

← All rights & provenance guides