Anthropic asks Claude users to share voice recordings for AI training

Anthropic has started asking Claude users to voluntarily share their voice conversations and audio recordings to help train its future artificial intelligence models.

The move, which is entirely opt-in, represents a push by the safety-focused AI developer to gather real-world speech data as tech companies compete to make their digital assistants sound more human.

How the voice sharing works

The request appears as a prompt when users start using Claude's voice features. It explicitly asks for permission to use the audio, stating: "Allow us to use your voice data to improve our AI models."

According to the text in the prompt, audio recordings and voice chat data help Anthropic's models improve how they understand and respond to speech.

Users are given two options on the screen: "Allow" or "Not now".

Because the setting is disabled by default, Anthropic is not automatically opting users into sharing their voice recordings. If a user chooses "Not now," their voice data remains private and will not be used to train the company's systems.

Managing privacy and data deletion

For those who want to change their mind later, Anthropic has added a dedicated toggle within the Privacy section of Claude's settings menu, labelled "Allow us to use your voice data."

The setting explains that, for approved sessions, users are allowing the use of their audio recordings and voice chat data to improve Anthropic's AI models.

The company says that users can turn the option off at any time. It also states that users can delete their existing voice data directly from the settings menu if they no longer want Anthropic to keep it.

Keeping voice separate from text

Notably, Anthropic is treating voice training as a completely separate preference from its existing data collection policies.

This separation means users have granular control over what they share. A user can allow Anthropic to train its models on their spoken conversations while keeping their written chats and Claude Code sessions private. Conversely, they can block voice training while still opting to share their text interactions.

The development comes as AI companies face scrutiny over how they source training data. By making the feature strictly opt-in and offering simple deletion controls, Anthropic is positioning itself as a more privacy-conscious alternative to competitors who have faced backlash for automatically training on user data.