Overview
IREX is an ethical AI platform for public safety. It turns an existing camera network, from a single site to an entire country, into searchable video intelligence. AI analytics detect threats as they happen, and investigation agents answer questions asked in plain language. Every privacy-sensitive action is bound to a case reference and written to an audit log that cannot be erased.
The platform is built and licensed by IrexAI Inc., a United States company. It serves public safety agencies, city and state governments, transportation authorities, campus and enterprise security teams, critical-infrastructure operators, and the system integrators who deliver for them. Through a partnership with the National Child Protection Task Force (NCPTF), IREX is also used to combat human and child trafficking and to locate missing children.
Two commitments shape every part of the design, and they are the reason this chapter exists before any procedure in this guide:
- Capability without compromise. Operators get real-time detection, sub-second search across petabytes of video, and natural-language investigations, at the scale of a national camera network.
- Accountability by architecture. The limits on that capability are enforced by the platform, not by policy documents. Recognition is restricted to pre-registered persons and vehicles, and permissions gate every screen. A Case ID gates every privacy-sensitive action, and the record of who did what is exported daily for independent review.
What the platform does
Four capabilities account for most of what operators do with IREX. Each is documented in its own chapter of this guide.
| Capability | What it means in a control room |
|---|---|
| Detect | 10+ AI video analytics suites run on live streams: faces against watchlists, license plates and vehicle attributes, weapons, fire and smoke, perimeter intrusion, crowding, railway hazards, traffic violations, camera tampering. See Key features and Video analytics. |
| Search | Searchveillance™ retrieves people, vehicles, and events from the archive in under a second, from a map, from filters, or from a photo. See Investigate a case. |
| Ask | Ask IREX interprets an investigation described in plain language and runs the multi-step search through the platform API. See Ask IREX. |
| Act | Alarm monitors raise real-time alerts; the Sover secure messenger carries them to the field, with evidence, so teams can coordinate a response. See Get alerted in real-time and Collaborate. |
Proven scale
The figures below are production numbers, not laboratory limits. They set the expectation for what a single deployment can carry.
| Measure | Figure |
|---|---|
| Cameras on a single instance, all running real-time analytics with 30-day instant search | 100,000+ cameras |
| Cameras managed across all deployments | 250,000+ cameras |
| AI video analytics suites | 10+ suites |
| Countries with production deployments | 8 countries: United States, United Kingdom, Ireland, Peru, Israel, Bosnia and Herzegovina, Mongolia, Kazakhstan |
Beyond the proven figure, the architecture scales linearly to 1,000,000 cameras, 100,000 users, and a one-year retention period without redesign. Growth in any one dimension is answered by adding nodes. See Platform architecture and Scale instance.
In this chapter
- Key features: what the platform provides, capability by capability.
- Platform architecture: how video travels from a camera to an operator, the deployment models, and the open-source foundation underneath.
- Video analytics: the analytics suites, and the operational scenarios they are built for.
- Ask IREX: natural-language queries and investigation agents, and the guardrails they operate inside.
- StreamVLM: custom detectors defined by a plain-text prompt, with no model training.
- Engineered for ethics: the six pillars of Transparency by Design, and the platform mechanism behind each one.
- Case ID and accountability: the mandatory justification gate, the Logbook, and the daily signed export that makes independent oversight possible.
- Data and technology sovereignty: who owns the data, where it runs, and what remains auditable.
- Regulatory alignment: how the platform supports compliance with CJIS, GDPR, the EU AI Act, the NIST AI Risk Management Framework, and US state privacy law.
Looking for a task, not a concept? The Quick start guide walks through the first four investigations end to end. Monitor, search, collaborate covers day-to-day operator work, and Set up, manage, comply covers administration.