Products — Cyber Defense AISOS
The defence instance
of the operating system.
This is not an endpoint product that grew an intelligence layer. It is AISOS, deployed to the place where execution happens — the same decision grammar, the same evidence contract, running where a process is about to do something irreversible.
Endpoint protection is one component of the system. It is not the system.
Measured
0.0233%
Measured false-positive rate
On a named clean corpus, after the calibration gate. An order of magnitude below the same engine measured before it.
0
Uncalibrated scores permitted to act alone
A model may recommend. Acting on a score with no corroboration is refused by policy, in the product.
µs
Decision latency at the endpoint
Judgement happens where the event happens. No network round trip stands between a process and being stopped.
3
Autonomy rungs, each revocable
Shadow, supervised, autonomous. The rung is a deployment decision, visible at all times and reversible at any point.
Measured internally on named corpora and re-measured on every release before it is allowed to ship. The corpora and the method are disclosed under engagement, not on a public page.
Engineering
Built as though it will be
attacked first.
A protection core that assumes it is the target
The integrity and protection core is written in Rust and built on the assumption that an adversary reaches it first. Self-protection is a property of the design, not a feature bolted to the side.
Decisions that survive the audit
Every enforced outcome carries the case, the corroborating signals, the policy that bounded it and the moment it was taken — reconstructable months later by someone who was not there.
Baselines that belong to one tenant
Behaviour is learned per institution. What is normal on a factory floor at 02:00 is not what is normal in a bonded customs system, and a shared baseline is how both get it wrong.
Containment that can be undone
Isolation, quarantine, process termination and rollback are bounded and reversible. A defence that cannot be safely wrong is a defence nobody will grant authority to.
Position
Do not add another dashboard.
Replace the layer that cannot act.
Where you already run tooling for visibility, reporting or compliance, this coexists with it. Where you run something that can only observe and escalate, it replaces it — and where you need the decision layer somewhere other than an endpoint, it goes there instead.
How it is fitted to an institution→