"Allow us to use your voice data to improve our AI models," Anthropic notes in the prompt.
Anthropic begins asking Claude users to share voice conversations
Anthropic has started asking Claude users to voluntarily share their voice conversations to help train and improve its AI models. The request appears as a new prompt when users engage Claude's voice features and explicitly asks them to allow Anthropic to use their voice data for AI training. According to the prompt text reproduced by the reporting, Anthropic tells users: "Audio recordings and voice chat data help improve how Anthropic AI models understand and respond to speech. Turn this off or delete this data anytime in settings."
How the Settings > Privacy option is worded
Anthropic surfaces the capability inside Settings > Privacy with a dedicated description: "Allow us to use your voice data." The setting further explains, "For approved sessions, allow the use of your audio recordings and voice chat data to improve Anthropic AI models." In the interface noted in the report, that voice-data setting is disabled by default.
The cyber insurance questionnaire just landed. Now what?
SOC 2, HIPAA, insurance renewals - someone has to own security strategy. Nubivance provides fractional CISO leadership without the full-time salary.
Get a security leadVoice training separated from chat and Claude Code training
Anthropic treats voice recordings as a distinct data category. Voice training is separated from the company's existing option to use regular chats and coding sessions for model training. That separation means a user can permit Anthropic to train on voice conversations while keeping regular chats and Claude Code sessions excluded — or make the opposite choice — using independent controls for each type of data.
User controls: opt-in toggle, "Not now," and data deletion
The company presents the voice-data request as optional: users can select either "Allow" or "Not now" when the prompt appears, and there is a dedicated toggle in Claude's Privacy settings. The setting seen in the reporting is switched off by default, so Anthropic is not automatically opting users in. Anthropic also says users can turn the option off later and that they can delete voice data from settings if they change their mind.
How technologists, procurement leaders, and end users will respond
- Technologists and security teams: They will note that voice data is managed separately from chat and code data, and that the platform exposes distinct controls and a default-off state for voice recordings.
- Procurement leaders and affected enterprises: They will be able to choose differing data-use permissions for voice versus text or code interactions because Anthropic's settings permit mixed choices across data types.
- End users and the general public: Individuals who value control over recordings can decline at prompt time, leave the voice setting disabled in Settings > Privacy, or delete any recorded audio later per Anthropic's stated options.
Anthropic's recent update is straightforward in its mechanics: a visible prompt, separate privacy toggles, a default-off setting in the interface observed, and an explicit deletion pathway. The combination gives users discrete choices about whether their voice conversations will be used to train models and allows different permissions for voice, chat, and code data. The reporting leaves one clear factual gap for observers to watch: whether and how many users will change the default and use the deletion controls Anthropic documents.
