- Observe Approved independent sensors and read-only data. Missing, stale or implausible inputs are flagged as lost visibility, never shown as normal.
- Evaluate Equipment-specific rules and local models compare readings with expected operating cycles, maintenance and conditions.
- Verify The responsible operator sees the readings, timing, model version and context. Physical risk is handled on its own merits, whatever the cause.
- Decide People interpret the evidence and decide whether to escalate through authorized procedures. Equipment is never switched by AI alone.
Existing protection and safety systems keep working independently and stay in charge. Urgent physical warnings go straight to staff.
Observe
Use approved independent sensors and read-only integrations where feasible. Identify stale, missing or implausible inputs.
Evaluate
Combine asset-specific rules with local models that account for operating cycles, maintenance and environmental conditions.
Verify
Show readings, timing, model version and context to the responsible operator. Separate physical risk from security attribution.
Respond and preserve
Support authorized escalation and separately engineered containment. Plan local evidence retention and controlled exports.