Anthropic adds opt-in voice-data sharing toggle for Claude model training

What happened: Anthropic has started prompting Claude users to voluntarily share their voice conversations (audio recordings and voice chat data) to train and improve its AI models. The prompt reads "Allow us to use your voice data to improve our AI models," with an Allow / Not now choice, plus a dedicated "Allow us to use your voice data" toggle under Settings > Privacy that is disabled by default. Anthropic says users can turn the option off later and delete voice data from settings.

Why it matters to creators: Voice is a biometric identifier at the core of voice-likeness rights (voice actors, podcasters, musicians). Anthropic's design — opt-in rather than opt-out, granular and separate from its existing chat/code training consent, disabled by default, and revocable — is a notable positive example of a consent-control pattern for voice data, and it signals that voice data is becoming routine training fodder.

Not yet known: Whether users are told how long shared voice data is retained, whether it is shared with third-party contractors or partners, and how Anthropic verifies consent for voices that appear in shared recordings but belong to someone other than the account holder. No legal analysis or liability assessment offered.