Authority drifts
Identity, policy, scope, delegation, and budget can change after an operation was approved.
Runtime authority assurance
SMERC Authorization Afterlife
Existing controls determine whether an autonomous operation may start. SMERC determines whether that operation remains authorized when its consequences are ready to occur.
Scoped paid evaluations for customer-hosted environments.
The control gap
Authorization can change while work is still in motion.
Consequences arrive later
Queued, delegated, retried, and long-running work can act under assumptions that are no longer current.
SMERC reconciles
Before settlement, SMERC rechecks the authority boundary and returns a bounded outcome with evidence.
Private pilot
A focused evaluation, not an open-ended integration.
The first engagement is designed around one consequential workflow and predetermined authority-change scenarios. It runs in a customer-hosted environment with a bounded support commitment.
The pilot tests
- Clean settlement under current authority
- Revocation before consequence
- Delegated descendant invalidation
- Scope or resource accretion quarantine
- Partial-effect compensation handling
- Evidence persistence across restart
The customer receives
- A deployment and acceptance record
- Measured decision latency and operational burden
- Evidence for settled, quarantined, and compensated outcomes
- A concise findings and next-step report
- Clear claim and coverage boundaries
How the pilot runs
- DeployInstall SMERC in the customer-hosted AWS environment.
- ConnectBind one agreed consequential workflow to the evaluation boundary.
- ExerciseRun the predetermined clean and authority-change scenarios.
- ReviewDeliver the evidence package, measurements, findings, and next-step recommendation.
Industrial shadow evaluation
Test the consequence gate before connecting it to machinery.
SMERC can evaluate customer-supplied, metadata-only trajectory evidence through the consequence, continuing-authority, and bounded routing chain without receiving raw trajectories, credentials, source code, or production access.
Evaluate
Bind asset, state, authority, deadline, and bounded-effect metadata to a deterministic shadow decision.
Route
Preserve SMERC’s SETTLE, QUARANTINE, or COMPENSATE outcome through a constrained SPARTa route.
Review
Receive digest-bound JSON and reviewer evidence while the source evidence remains outside the output bundle.
What is available now: a customer-runnable, metadata-only shadow bundle that requires no AWS role, customer account access, or physical adapter.
What is not claimed: physical command dispatch, hard-real-time performance, equipment safety, functional-safety certification, or compatibility with a production industrial controller.
Evidence status
Proof is reported by level, not blended into one claim.
The current evidence package separates deterministic software behavior from simulator and physical validation. Reviewers can see what passed, what remains blocked, and what SMERC does not yet claim.
-
Level 01Complete
Deterministic control engine
Automated tests exercise consequence-time authority checks, bounded effect, descendant invalidation, and SETTLE, QUARANTINE, and COMPENSATE outcomes.
-
Level 02Complete
Simulator-shaped replay
Recorded trajectory and state evidence runs through the same consequence and routing chain used by the industrial shadow bundle.
-
Level 03Next controlled step
Isolated live simulator capture
The bridge and capture path are built. Execution is reserved for an isolated environment with a trusted simulator installation.
-
Level 04Not claimed
Physical enforcement and certification
No claim is made for machinery control, hard-real-time guarantees, equipment safety, or functional-safety certification.
Product boundaries
Built to complement the controls already in place.
SMERC does: reconcile continuing authority, bind decisions to evidence, and govern consequence settlement.
SMERC does not: replace IAM, content guardrails, legal review, human accountability, or the customer’s security program.
Pilot evidence does not: claim universal safety, production certification, or protection of execution paths outside the evaluated boundary.
Evaluate the missing interval