Responsible AI
Honest by policy.
In fire safety, an overclaimed detector is more dangerous than none. We engineered the claims down and the human oversight up.
Commitments
How we keep the AI safe
Human verification is mandatory
Every AI event is advisory. A trained person reviews it before any action or closure.
False alarms need a reason
Dismissing an event requires a documented reason — the feedback loop that keeps the system honest.
No overclaiming
No certified-detection promises, no alarm replacement, no accuracy figures we have not measured on your site.
Smoke/flame deferred
Visual smoke and flame detection is on the roadmap only after proper validation — and even then, advisory.
Transparent events
Each event carries a plain-language rule explanation, so reviewers understand why it was raised.
Privacy-preserving
Fire Watch uses no facial recognition and no identity analytics; on-premise by default.
Safety positioning. Leal Vision AI — Fire Watch is an assistive visual analytics system. It does not replace certified fire alarm systems, statutory smoke, heat, or flame detectors, fire panels, emergency procedures, or trained safety personnel. All AI-generated events are advisory and require verification by your safety team.