XS1 // CYBER

Secure the environment.

The cyber research and engineering division of XS1. Defensive systems, secure AI, and adversarial capability studied under lab control — so defenders see the attack path before an adversary walks it.

Project North Star Discuss a Cyber Program

xs1-cyber // chain walkthrough

operator@xs1:~$ xs1 recon --scope authorized --depth full

█

Illustrative walkthrough — authorized engagements only

// zero-day discovery// exploit chaining// proof of exploit// secure ai// detection engineering// containment// adversary emulation// after-action evidence// zero-day discovery// exploit chaining// proof of exploit// secure ai// detection engineering// containment// adversary emulation// after-action evidence

01 / Division capability

Research and operations, in the same division

01Defensive systems

Detection, hardening, and response engineered into the systems they protect rather than bolted beside them.

› detection engineering
› hardening baselines
› response runbooks

02Autonomous red-teaming

Continuous adversarial pressure from agents that hold state across a campaign and never leave authorized scope.

› stateful campaigns
› scope guards
› operator approval gates

03Zero-day discovery research

Reasoning from a specific configuration to a specific, demonstrable consequence — not a list of theoretical risk.

› config-to-consequence
› reachability analysis
› reproducible findings

04Exploit chaining

Composing individually minor findings into the path an operator would actually walk, with each hop evidenced.

› path composition
› precondition proof
› impact modelling

05Secure AI infrastructure

Agent systems that can be trusted with privileged access, sensitive data, and irreversible actions.

› least privilege
› isolated execution
› audit trails

06Cyber operations

Standing capability: monitoring, escalation, containment, and after-action evidence a reviewer can follow.

› monitoring
› escalation
› containment

02 / Flagship program

Project North Star

A long-horizon program to build Polaris — a 300B-parameter model specialized in offensive security reasoning. Currently in post-training, with closed weights and no announced release.

Enter the program page

Model

Polaris

Parameters

300B

Stage

Post-training

Weights

Closed

XS1 // CYBER

Secure the environment.

Cybersecurity, defensive systems, secure AI, adversarial research, and cyber operations.

Discuss a Cyber Program General contact