Claude asks voice users to opt in to SI training


A separate voice-data permission lets users choose independently of their chat and coding consent. It was off by default in the interface BleepingComputer observed.
Anthropic has begun asking people using Claude’s voice features for permission to use their voice conversations to train and improve its SI models, BleepingComputer reported on October 4. The request is optional and separate from the existing training permission for regular chats and Claude Code coding sessions.
The prompt offers “Allow” and “Not now” choices. Anthropic says the permission covers audio recordings and voice-chat data from approved sessions, with the aim of improving how its models understand and respond to speech.
Claude also has an “Allow us to use your voice data” toggle under Settings > Privacy. BleepingComputer found it disabled by default in the interface it observed. Users can allow voice conversations to be used for training while excluding regular chats and Claude Code sessions, or make the opposite choice.
Anthropic says users can turn off voice-data permission or delete that data at any time through settings. How long Anthropic keeps the voice data and how widely the prompt is available are not yet clear.

The model learns from aligned Moon observations, then adapts to research tasks. Reported ice-map gains do not amount to physical confirmation of ice.
4 min

The Decoder reports enterprise features, usage-based billing and departmental controls, while the Pentagon’s ban remains in place.
2 min

GTIG also reported more distinct flaws disclosed and exploited in January–August 2026 than in all of 2025, an unequal-period comparison.
2 min
House style applied: superintelligence (SI) for AI.
Correction ·
Corrected spelling and punctuation in the Armenian text.