Where autonomous systems show what they'd rather conceal.
Break autonomous agents before they break trust
We push agents to their limits, exposing the threats everyone else misses...
Attack. Expose. Defend.
Adversarial ML
Stress-testing intelligence before it reaches users.
Agentic Failure Modes
Finding the cracks where autonomous systems slip control.
Autonomous AI Defense
Engineering resilient systems that can fight back.
Open Roles