Mate Labs publishes the Blue City benchmark, without Mate in it

11-03month and day

What happened

Mate Labs published Blue City on 2026-11-03: a simulated enterprise used to score AI defenders. Five out-of-the-box agents were each run on 32 scored incidents. The page reports effectiveness from 40.2 to 80.1 percent, cost from 0.45 to 4.05 USD per incident, and traceability from 0.0 to 3.1 percent. The method is published in full, with eight scenario families, twin pairs, a damage formula and a cost model. Page read 2026-11-03.

Mate's own system is not among the five agents. The page says the benchmark is not validated against real enterprises, the damage weights are unpublished, and no outside party has reviewed it. This is a vendor-run test of other products, and it measures investigation and damage, not time to contain.

What it changes for a buyer

It is a useful template for how a method can be written down, and a reminder that a published method is not a result for the vendor that wrote it. It does not change Mate's scores here, because it does not test Mate and does not time a containment.

All notes are on the updates page. The scores on this site change only when a vendor-controlled page changes; see the corrections log.