Watermarking & the audit log

MeetCrew marks its output and keeps its compliance record durable. Synthetic speech is watermarked, generated artifacts carry embedded origin metadata, and every compliance event is written to a tamper-evident log.

Marked output

Every segment of synthesized speech a teammate produces carries an inaudible audio watermark, sealed at the moment of synthesis so the mark stays inside MeetCrew’s trust boundary. The banner and the watermark are applied fail-closed: a video frame without its banner is never sent, and a speech segment without its watermark is never sent.

Origin metadata on generated artifacts

Visual artifacts MeetCrew generates — a teammate’s profile avatar, a composited meeting background — carry embedded XMP origin metadata recording what was made, when, and by which teammate. It is readable by any standard metadata tool, so you can check it yourself rather than take our word for it.

Two limits, stated plainly. It is not a signature: the metadata records origin, and it is not tamper-evident, so it cannot prove an artifact is unaltered. And it covers the images MeetCrew produces — not transcripts, recordings, or exports, which carry no embedded credentials today. An image you upload is recorded as received rather than as authored by us, and one that already carries its own provenance metadata is left exactly as it is.

The audit log

Disclosures, opt-outs, and deletions are written to a hash-chained, tamper-evident audit log. Each entry links to the one before it, so any alteration breaks the chain. The chain is re-verified nightly, and it is fail-closed: a broken link raises an alert and is never silently overwritten. The content of an opted-out meeting is gone, but the fact that the opt-out happened — with its timestamp and the invoking speaker — is retained.