CrowdStrike just rolled out SafeMind—a next-gen agent-based cybersecurity system built in partnership with Nvidia. The announcement dropped on September 1 at the Fal.Con 2026 conference in Las Vegas. SafeMind is set to run natively on the CrowdStrike Falcon platform.

What is SafeMind?

SafeMind brings together offensive and defensive AI models in a closed-loop setup. One model actively looks for ways to attack your infrastructure, while the other creates and tests rules to spot those attacks. Think of it as a non-stop chess match between an attacker and a defender, all inside one system.

How it Works

The offensive side generates possible compromise paths, and then the defensive model builds detection rules and validates them against the same data. This loop lets teams quickly test out new attack scenarios and instantly back them up with fresh detection rules in the Falcon environment.

Models at Launch

At launch, there are two specialized models: Red Tempest simulates the attacker, searching for potential compromise vectors, while Blue Solano analyzes the data to detect threats as they emerge.

Where and When It Was Announced

SafeMind was officially announced on September 1 at Fal.Con 2026 in Las Vegas and is being positioned as a native part of the CrowdStrike Falcon platform.