A separate prompt in Claude was reported in early October 2026, asking users to let Anthropic use audio recordings and voice-chat data from approved sessions to improve its AI models. The reported interface offered a choice to allow that use or defer it, and showed a separate voice-data toggle in Settings > Privacy.
What the reported prompt asks
The reported setting concerns Anthropic using eligible voice data to improve its models. It is a choice about Anthropic’s model improvement, not a tool for users to train a personal voice model or fine-tune Claude. The voice-data toggle was off by default in the interface described.
A separate choice from chats and coding sessions
The reported voice-data control is separate from the existing setting for chats and coding sessions. Anthropic’s consumer policy says those sessions may be used to improve its models when users allow it, in specified safety-review circumstances, or through another explicit opt-in.
Voice Mode and dictation handle speech differently
Anthropic describes Voice Mode as a two-way spoken conversation; its text transcripts are saved in chat history. Mobile dictation, by contrast, converts speech into text. Anthropic says dictation audio is deleted after transcription and that voice is not used for model training under that policy. Transcribed text may be used according to the user’s chat settings.
What users can change
The reported control lets users switch off the voice-data setting later or delete voice data in settings.
