Solutions & Use Cases

No, and it should not be evaluated as one. Those systems execute door control, mass notification and dispatch. This provides the visual intelligence layer that can send validated events to them through integrations -- REST APIs, webhooks and the workflow node catalog -- rather than replacing them. There are no proprietary lock or notification systems here to buy.

It is not designed to. Response runs through workflows the institution defines, with human approval gating consequential steps and escalation rules set by policy, not by the model. An AI-raised event can reach an external system only through the integrations you configure, under the controls you set.

No system can reliably detect bullying, and claims otherwise deserve suspicion. What this detects is narrower and observable: aggressive physical interaction, crowding, unusual activity and scene context that helps a person assess what deserves attention. It is a triage aid for staff, not an autonomous disciplinary decision-maker.

Detection models watch for brandished weapons on the live streams and flag the frame the moment one appears. The flag goes to a person for verification before anything escalates, and confidence thresholds are configurable per camera. Detection accuracy gates the whole chain, so the verification step is a design feature, not a disclaimer.

The detection is -- it reads the live stream and fires as the condition occurs. The added scene description runs on cached frames on a cadence, so it arrives moments after rather than in the same instant, and the alert does not wait for it.

Role-based access controls who sees footage, retention policy governs how long it exists, every access is audited, and redaction removes faces and identifying detail from any copy that leaves -- the original stays intact. Deployment can be on-premises or fully air-gapped where footage must not leave the institution, and school video handled as an education record stays inside those controls.

The footage around the event is already preserved with chain of custody. Investigators search the estate in plain language, assemble what happened across cameras, and produce a record that survives a district review, a parent meeting or a legal process -- with a redacted copy for anything disclosed.

No. Detection runs over RTSP and ONVIF against the existing estate, alongside whatever VMS or NVR is already in place, with camera-health alerts so a dead camera is found before it matters.

Per-user tokenized URLs with time limits and access-count restrictions, domain whitelists that control where content can be embedded, dynamic watermarking that traces every copy to its recipient, and audit logging of who opened what, when, from where.

Yes. 33+ spoken and 40+ total PII categories are removed before any external eye sees the file. The redacted copy travels; the source record never leaves custody.

Yes. Secure video sharing with HIPAA, GDPR, and CCPA-supported configurations, so claims recordings, patient media, and client files move to processors and partners without a compliance exception.

Paired with DEMS, disclosure moves through prosecutor and defense portals with tokenized, expiring external links, and the chain of custody stays intact across the exchange.

Links expire by time or view count, embedding is walled to approved domains, and the audit log shows exactly who had access during the window.

No. Your policy sorts every action into what runs automatically, what waits for a named person, and what is never automated. A person can always act without the system.

It is not part of the offer. The software detects events you can see, like a weapon being drawn, a person on the ground or a crowd building. Where the law requires identification workflows they are handled separately under their own governance, never by default.

Yes. It connects to the cameras and video systems you already run using the standard streams they publish, so there is no replacement program. Each camera gets tuned to its own view during setup, and cameras with built in analytics feed the same alert queue.

You decide how each kind of event is handled. Uncertain detections are checked by the AI before anyone is paged. Serious ones page a person right away, with the AI review arriving moments later. The AI can raise urgency but it can never quietly cancel an alert, and nothing irreversible happens without a person.

On your own servers, in a government cloud, or a mix of the two. For live video analysis most cities run it on premises, and footage never has to leave your control.

Yes. Camera zones give you counts, occupancy and dwell for each view. They are operating signals, good for detection and trends. Where a certified study is required, they tell you where to point the instruments.

Every camera is tuned to its own view against a recorded baseline, weather and headlights included, and uncertain detections get an AI review before anyone is paged. The first stretch runs quietly with nobody alerted while thresholds settle on your actual conditions.