We build World Models for cybersecurity.
Harbinger Labs is an early-stage research company. We are building models that understand an organization well enough to predict how an attacker will move through it, before they move.
Who we are
We started Harbinger Labs on one conviction: the attacker is now learning, so the defender has to foresee.
What we are building
Our first model is Scry. It reads a network from a single export, plays every intrusion forward thousands of times, ranks the ones that would land unseen, and names the one change that stops them. Then it does it again tomorrow.
- 01Attack-path models
Rank where an intrusion moves next, on networks the model has never seen.
- 02World Models
Predict the outcome of each move, and the effect of each fix, before anything is touched.
- 03Blind-spot models
Surface where your telemetry goes dark. The quiet route is the dangerous one.
We begin with identity, because it sits at the center of the modern enterprise. The same model extends to cloud, infrastructure, and the physical world.
Every breach was once only a possibility.
We end it there.