IncidentScribe for defence & aerospace
In defence and aerospace environments, the rule isn’t “be careful with the cloud tool” — it’s “the data does not leave.” Incident artefacts can carry controlled or sensitive information, and an outbound call to a commercial LLM API is exactly the kind of egress these environments exist to prevent. The entire category of cloud incident-AI tools is simply unavailable to you.
The gap that leaves
Incidents still need rigorous, documented reviews — arguably more so here than anywhere. Without an approved tool, that work falls back to manual drafting, with all the postmortem debt and inconsistency that comes with doing it by hand under deadline pressure.
An air-gap-friendly shape
IncidentScribe runs Apple’s Foundation Models fully on-device. The drafting workload makes no network call — the model is local to the Apple Silicon Mac. Outbound access is limited to StoreKit receipt validation and enforced by the macOS App Sandbox; there is no telemetry, no remote logging, and no third-party SDK. The incident data and the drafted postmortem stay on the machine by construction, which is the only posture that works when egress is the threat model.
Traceable for the review board
Every drafted statement resolves back to its source artefact through the citation chain; events that can’t be grounded in the source are dropped during validation. The output is a postmortem a review board can audit line by line — produced without a single byte crossing the boundary.