AI news · Anthropic · privacy
Claude asks users to share voice recordings for training
A new opt-in prompt and its own privacy toggle cover audio from Claude’s voice mode, apart from the chat training setting.
Anthropic has started asking Claude users for permission to train its models on their voice conversations, BleepingComputer reported on Sunday 4 October 2026. A prompt in Claude’s voice mode offers Allow or Not now, and a matching control sits under Settings, then Privacy.
The control is off by default in the version both outlets saw. It covers audio recordings and voice chat data from sessions the user approves, and Anthropic’s text says the user can turn it off or delete the data at any time. Chats and Claude Code sessions keep their own training switch, and either one can be on while the other stays off.
gHacks points to two gaps: no retention period appears in the prompt or the setting, and Anthropic has not said how widely the request is rolling out.
For anyone who uses Claude’s voice mode with client material, the new toggle needs its own check: switching off chat training leaves the voice setting as it is.
Sources
- BleepingComputer: Anthropic asks Claude users to share voice data for AI model training. Checked
- gHacks: Anthropic Asks Claude Users to Share Voice Recordings for AI Training With a New Opt-In Setting. Checked