A machine runs a command only if it is the one an operator approved.

Last updated: 15 November 2026

A PLC, a robot controller or a SCADA endpoint executes any well‑formed command from whoever can reach it. A compromised engineering workstation, a stolen credential, or an SI agent (Super Intelligence, our term for what the industry calls artificial intelligence) acting on a poisoned instruction can send a perfectly valid command that nobody approved.

SPT‑Txn applies the same check it applies to payments: authority exists only inside a short‑lived token bound to one approved action, and the command is refused if it is not that action.

How it works

Built for the air gap

Verification is offline. The gate checks the token against a locally held, signed trust bundle, so nothing needs to call out of the plant network and nothing calls home to us. Revocation travels inside the bundle, and each bundle carries a hard expiry window. The trust‑bundle format is under review.

We authorize. We are never the safety function.

SPT‑Txn decides whether a command was authorized. It does not replace interlocks, safety instrumented systems or any protection layer, and nothing safety‑critical should ever depend on it. It sits on the command path, in front of what already keeps the plant safe.

Honest boundaries

This is demonstrated in simulation, not on plant hardware. There is no integration with any controller vendor today, and nothing here has been externally audited or certified against IEC 62443 or any other standard. It is not in production use. The same engine runs our payment demos; the industrial profile is the newest part of the work.

If this is useful to you

We are looking for a plant, lab or integrator willing to test the gate on real equipment.

The same check for drones (design) →
The same check for robots and IoT (design) →
How the token chain works →
Reference implementation on GitHub →
rudi@violetskysecurity.com