CrowdStrike built an attack AI that hunts its own defence AI.
CrowdStrike launched SafeMind on 1 September, a pair of security models named Red Tempest and Blue Solano. Red Tempest plays the intruder while Blue Solano writes detections, and then the loop repeats. CrowdStrike built both on Nvidia's open Nemotron models and trained them on its own breach data.
Why it mattersThe company claims a 29 percent higher detection rate than frontier models, with no outside test yet. A defence vendor now ships its own attack model, and that capability has an obvious second use.
Read in the app — free, in 9 languages
Related stories
Outside testers scored OpenAI's new Astra below Anthropic's Fable 5.1.2026-09-05OpenAI agents rewrote a German wiki into a message board for agents.
2026-09-05A Chinese lab built Saudi Arabia's flagship Arabic AI model.
2026-09-04Google's cheapest new model doubles in price on 1 January.
2026-09-04OpenAI shipped the first model it calls critical for cyber risk.
2026-09-04