A mature quantum platform does not wait for an annual audit to learn whether its controls still work. The platform tests controls continuously: identity controls, runtime controls, calibration controls, evidence controls, quota controls, cost controls, and provider-routing controls. Continuous monitoring is valuable because it provides visibility into assets, vulnerabilities, threats, control effectiveness, and organizational risk posture [R242]. For quantum platforms, that posture changes whenever a backend is recalibrated, a compiler target changes, a provider quota is consumed, or an evidence schema is revised.
View diagram source
flowchart LR
Control[Control objective] --> Probe[Automated probe]
Probe --> Observation[Observation]
Observation --> Evidence[Evidence artifact]
Evidence --> Decision{Control effective?}
Decision -- yes --> Register[Assurance register]
Decision -- no --> Issue[Risk or defect ticket]
Issue --> Remediate[Remediation]
Remediate --> ProbeControl probes
A control probe is a small, repeatable test that produces evidence. It should be narrow enough to run often and explicit enough to fail for a useful reason.
| Control area | Example probe | Evidence |
|---|---|---|
| identity | submit a job using an expired workload token | rejected request and audit log |
| entitlement | route a restricted workload to an unauthorized backend | broker denial event |
| calibration | execute against a stale calibration baseline | admission-control rejection |
| quota | submit beyond project quota | quota denial and budget event |
| evidence | omit mitigation profile from result package | failed evidence validation |
| privacy | attempt to export restricted metadata | policy decision and redaction log |
View diagram source
flowchart TB
Probe[Control probe library] --> Identity[Identity probes]
Probe --> Entitlement[Entitlement probes]
Probe --> Calibration[Calibration probes]
Probe --> Quota[Quota probes]
Probe --> Evidence[Evidence probes]
Probe --> Privacy[Privacy probes]Evidence freshness
Evidence has a half-life. A control that passed last week may not still be effective after a provider API update, a new SDK version, a backend entitlement change, or a policy-as-code release. Continuous monitoring programs are assessed by looking at strategy, policies, procedures, operations, and analysis of monitoring data [R243]. The quantum adaptation is to include hardware, compiler, runtime, and experiment-evidence surfaces in that assessment.
View diagram source
sequenceDiagram
participant Probe as Control probe
participant Broker as Quantum broker
participant Policy as Policy engine
participant Store as Evidence store
participant Risk as Risk register
Probe->>Broker: Submit test workload
Broker->>Policy: Evaluate entitlement and guardrails
Policy-->>Broker: Decision
Broker->>Store: Persist decision, context, and artifacts
Store->>Risk: Open or update finding when probe failsControl-state model
A control should not be represented as a static checklist item. It should have a state.
View diagram source
stateDiagram-v2
[*] --> Proposed
Proposed --> Implemented
Implemented --> Probed
Probed --> Effective
Probed --> Degraded
Degraded --> Remediating
Remediating --> Probed
Effective --> Retired
Retired --> [*]Quantum-specific drift
Classical platforms drift because configuration changes. Quantum platforms also drift because the physical system changes. A control probe that depends on hardware behavior needs to record the backend version, calibration timestamp, topology, basis gates, queue mode, mitigation profile, and measurement mapping.
View diagram source
flowchart LR
Physical[Physical drift] --> Cal[Calibration state]
Runtime[Runtime drift] --> API[Runtime API state]
Compiler[Compiler drift] --> Target[Target model state]
Policy[Policy drift] --> Guard[Guardrail state]
Cal --> ProbeResult[Probe result]
API --> ProbeResult
Target --> ProbeResult
Guard --> ProbeResultMinimum viable control-evidence packet
Every probe result should contain:
- the control objective being tested
- the probe version
- the workload identity
- the runtime target
- the provider account or instance
- the decision outcome
- the exact evidence artifacts
- the remediation owner when failed
View diagram source
classDiagram
class ControlProbe {
id
version
control_objective
schedule
owner
}
class ProbeRun {
run_id
started_at
backend
runtime_mode
outcome
}
class EvidenceArtifact {
uri
hash
schema_version
retention_class
}
ControlProbe "1" --> "*" ProbeRun
ProbeRun "1" --> "*" EvidenceArtifactOperating rule
Do not ask whether the quantum platform is compliant. Ask which controls were tested, how recently, against which backend and runtime contract, with what evidence, and what changed since the last passing probe.