Deep Hat by Kindo — Product Research | aVenture Research
Deep Hat is Kindo's proprietary security model descended from WhiteRabbitNeo, an open source model exploring real offensive and defensive tasks with language models, serving autonomous SecOps, DevOps and ITOps workflows. Trained on real attack patterns and operator tasks, it supports exploit exploration, adversary emulation and multi-step offensive reasoning. Sift skips queued events an agent would dismiss once a person approves the threshold; Shrink runs repeatable workflow steps as code, routing exceptions back to the model.
Deep Hat executes inside Kindo, keeping the work in the customer's environment with no external calls and no exposure of sensitive data, unlike censored models that refuse much of the offensive and investigative security work operators need done. It works with a customer's existing CLI, shell and security stack for security tasks that need privacy and precision.