Anthropic Encourages Users to Share Voice Data for AI Model Improvement

Anthropic, the San Francisco-based AI company, is prompting users to voluntarily share their voice conversations to improve its AI models, including Claude Opus and Claude Fable. This request comes amid ongoing discussions about how AI companies gather training data.

The core of Anthropic's request is straightforward: users can opt in to share their voice conversations, which will be utilized for model enhancement. Previously, the company’s voice features operated differently, with audio recordings deleted after transcription and not used for training purposes. This change follows a significant update to its consumer terms on August 28, 2025, allowing user consent for data usage in model improvement.

Under the new terms, users on the Claude Free, Pro, and Max plans can choose to share their data, with the option to opt out at any time. Users who opt in will have their data retained for up to five years, while those who opt out will revert to the previous arrangement of 30-day data retention. Notably, the default setting for new users is to have the “Help improve Claude” toggle switched on.

This initiative positions Anthropic competitively within the AI landscape, as it seeks to enhance its models' speech recognition and response capabilities. The company emphasizes user interactions as a means to improve both the functionality and safety of its AI systems.

As the AI sector evolves, market observers will be keen to track the opt-in rates and user sentiment regarding this new data-sharing approach, as well as its impact on Anthropic's standing against competitors like OpenAI and Google.