We’re looking to build an internal tool on the Meeting SDK and want to confirm it’s compliant before we start, given the docs say “reserved for human use cases, no bots or AI notetakers.”
The setup:
A real, signed-in host/panelist joins their own account’s webinar through our custom Meeting SDK app
The app reads incoming attendee chat questions and shows the operator a suggested draft reply (the draft is AI-assisted)
A human reviews each suggestion and manually clicks send. Nothing is ever sent automatically — a person approves and sends every individual message, using ‘SendChat4WebinarMeeting’ to reply privately to the attendee who asked
So there’s no unattended bot and no autonomous sending. A present human operator is in control of and sends every message.
Questions:
Is this human-in-the-loop pattern (human reviews AI-suggested text and manually sends each message) compliant with the Meeting SDK’s “human use cases” policy?
Any requirements we should follow to stay compliant (attribution, disclosure/consent to attendees, ZAK vs OBF since it’s our own account’s webinars, etc.)?
Zoom does not explicitly confirm that AI-assisted drafting is compliant, so the published guidance on human use cases does not prove a definitive answer.
Implement every applicable Legal UI Notice and Active App Notifier requirement, and disclose in your privacy materials if attendee chat is sent to an external AI provider. Before development, submit this exact data flow and manual-send design through Zoom’s review/support channel and request written confirmation that it remains a human use case.
For managed participant-based integrations, Recall.ai’s Meeting Bot API may also be useful.
@Gavin1 yes this is considered as human use case, as a real human is using and operating the meeting sdk. You will sign in to the meeting using ZAK token
