Recording processing permission and retry

Only the recording owner can start transcription, regenerate its title, summary or actions, or recover a bot transcript. A workspace role or a shared link does not authorize someone else's processing.

1. Open Profile → AI sharing → Review AI sharing. Select Summaries and extraction with Google Gemini and its Allow background processing for this service checkbox. For transcription, also select each needed Transcription with Groq, Transcription with AssemblyAI or Transcription with Deepgram service and its background checkbox. Search indexing with Google Gemini is a separate optional choice.
2. Choose Allow AI sharing, then reopen your recording. Use the existing Regenerate control for its title, summary, actions or transcript. On a waiting bot transcript, choose Check now or Transcribe now; Meeting capture with Recall also needs its own background checkbox. Fallback only has access to recipients you explicitly allowed.
3. Wait for the saved result. Queue acceptance is not completion. If permission is refused, changed or cannot be verified, review your choices before explicitly trying again. An uncertain request is never automatically sent twice; check the recording first.

Saved media remains available when processing permission is refused. Cancelling the local wait does not confirm that submitted provider processing stopped. The waiting screen reads existing results without silently starting a new fallback job. Changing choices does not resume an old queued job or prove that an external job/bot stopped. Scheduled AI briefings, automated coaching and external assistant relays are currently unavailable; background permission does not enable them. AI support requires the separate per-request approval described below. Microphone permission, feature access, billing and participant/import rights remain separate.