Push-to-talk first
Explicit activation keeps the first voice experience predictable and privacy-forward.
Hands controlled first
Voice Experience
Voice should feel alive while respecting privacy, consent, retention, and visible microphone state.
Transcript parity
Users see what was heard and can correct it before sensitive work proceeds.
Barge-in
Interrupt, stop, and redirect without silent continuation of background actions.
Retention controls
Audio retention should default off, with clear policy and enterprise control options.
Safe speaking
Sensitive content can be muted or restricted where policy requires.
Same governance
Voice input still becomes a governed request with route, policy, and trace evidence.