About

Real-world defense takes people who have been on offense

Models are learning to attack far faster than they are learning to defend. We build the training infrastructure that closes the gap.

Offense trains on a clean signal. The flag drops or it does not, a script checks it, and the model improves with no human in the room. That loop scales, and it is why offensive capability is compounding fastest.

Defense has no equivalent. A defender has to stop the attacker and keep the mission running, and those are two different outcomes. Isolating a compromised host ends the intrusion and takes the business down with it. Today that judgment costs a human every single run, so defensive training does not scale the way offensive training does.

Existing cyber training does not close the gap. Detection work asks whether an event is malicious. Exploit and patch gyms ask whether a bug reproduces. Neither asks the question that matters under a live adversary: which action breaks this attack chain without breaking the system it is defending?

SentryStack builds the environments, platforms, tools and verifiers that answer it. Each environment is a wargame: a real AI-enabled attack chain, authored by operators who spent their careers building ranges for offensive testing, replayed against a defender on reproducible infrastructure. A privileged verifier watches attacker progress, mission uptime, collateral impact and remaining footholds, and scores the intermediate steps rather than handing the judgment to another model.

The result is a defensive reward that a machine can check: attack stopped, mission intact. That is what makes defense trainable at scale, and it is the piece nobody has built yet.

Attack stopped. Mission intact.

Founders

Connor, Co-Founder & CEO of SentryStack

Connor

Co-Founder & CEO

Peter, Co-Founder of SentryStack

Peter

Co-Founder

Technical Advisors

M

Max

Technical Advisor

FDE @Palantir

N

Nathan

Technical Advisor

FDE @Palantir

Want to follow along?

We're talking to frontier labs, defense programs and operators who want to work on defensive RL. We recruit from the same community we came from.