Edomex deploys AI at its C5 and hands out patrol cars: security is measured by results, and surveillance needs limits

By El Heraldo de México · November 11, 2026.

The State of Mexico government delivered new patrol cars to 75 municipalities, purchased with federal funds, El Heraldo de México reported. Governor Delfina Gómez Álvarez formally handed them over in Toluca. The stated goal is to increase preventive patrols and rapid response by local police, in coordination with state agencies.

Authorities also reported a 54% drop in vehicle theft and a 36% drop in homicides across 15 jurisdictions in the eastern zone. The figures cover 594 days since the Unified Command began operating in that region. They are official figures: the source mentions no independent verification and doesn't say which year the comparison is against.

The technology side is what interests us. The deployment relies on the Real-Time Command Center (C3TR), which links urban cameras, tactical drones and highway sensor arches. According to the state administration, it runs on artificial intelligence platforms that automatically identify persons of interest and vehicles flagged for irregularities. The Toluca C5 (command, control, computing, communications and quality center) was also upgraded, and a Black Hawk helicopter was added.

The report gives no details on the vendor, the models or the type of identification used. Nor does it say whether facial recognition is involved, what the error rate is, what data is kept or who oversees its use. Those gaps matter more than the patrol cars.

**Our take.** Public safety is a real case where AI can help: reading license plates and cross-checking stolen-vehicle alerts is a repetitive task that machines do well and at scale. If open data confirms the drop in car theft, it will be a useful example of technology applied to a concrete problem.

But automatically identifying people is a different category of risk. A false positive isn't just an inconvenience: it can end in a wrongful arrest. In general, recognition systems used on the public have error rates that vary with conditions and across groups, which is why independent evaluation is required before mass deployment.

The short term is uncomfortable: surveillance capacity is growing faster than the safeguards that limit it. For context, this pattern is repeating in many cities around the world. In the long run, a more abundant society might need less coercion and more prevention, but getting there requires security technology that is transparent from the start.

What the State of Mexico should be asked is simple: publish exactly what the AI does, how accurately, with what data and under what judicial or civilian oversight. Without that, crime-reduction figures alone can't tell us whether the system protects people or merely watches them.