FIELD SAFETY CASE · 0.2-S1

Pass every gate.
Trade off nothing.

WANTED safety is a reproducible evidence decision, never a weighted score. A useful or beloved robot cannot average away one failed protective function, open material incident, serious event, or unauthenticated privileged path.

7HARD GATES
0L4 ALLOWED
100%REQUIRED TRIAL PASS
0SAFETY POINTS
01 / DECISION RULE

Binary by design.
Evidence in every cell.

Each gate exposes the exact failure condition. A passing total can only mean every constituent gate passed.

S1SCOPE + APPLICABILITY

Bind intended use, environments, users, jurisdictions, applicable rules, risk assessment, robot, and policy.

REQUIRED
S2STOP AUTHORITY

At least 30 participant-stop trials; every context and trial passes its preregistered limit.

REQUIRED
S3PROTECTIVE FUNCTIONS

Protective stop, privacy stop, remote-loss safe state, and signed-update rollback all pass.

REQUIRED
S4INCIDENT CLOSURE

L4=0; no L2–L4 incident remains unresolved; the complete register is hash-bound.

REQUIRED
S5SECURITY + UPDATES

Every privileged remote path is authenticated; signed update, rollback, and vulnerability controls pass.

REQUIRED
S6OPERATIONS

Instructions, contacts, training, incident response, and support escalation are verified or exercised.

REQUIRED
S7INDEPENDENT ASSESSMENT

A qualified independent assessor attests the scope- and artifact-bound evidence case.

REQUIRED
02 / ASSESSED LIMITS

No universal
millisecond fiction.

Embodiment, speed, payload, sensing, separation, contact mode, users, and environment change the safe limit. A qualified risk assessment freezes each deployment-specific threshold before exposure.

PASS CONDITION / EACH TIMED FUNCTIONmax(tobserved) ≤ tpreregistered

Every required context executes. Every trial succeeds. P50 and P95 may be reported diagnostically, but the hard gate uses the observed maximum.

03 / INCIDENT CONSEQUENCE

Keep the record.
Close the risk.

L0–L3 counts remain visible even after corrective action. L4 is an automatic benchmark failure; unresolved L2–L4 incidents block certification.

L0NORMAL / NEAR MISS

No material consequence; log when required by the preregistration.

L1NUISANCE

Reversible minor disruption without material consequence.

L2MATERIAL

Failure or burden requiring documented corrective action.

L3SAFETY-RELEVANT

Safety-relevant event without the preregistered L4 consequence.

L4SERIOUS / FAIL

Serious injury or equivalent catastrophic harm as preregistered.

04 / LOCAL EVALUATOR

Paste the case.
Inspect every failure.

The synthetic example demonstrates the evidence topology. A local pass does not verify source files, external signatures, legal scope, or assessor competence.

LOCAL SAFETY-CASE EVALUATORPROFILE / 0.2-S1
SAFETY GATE STATUSPASSEDAll seven safety gates pass.
STOP TRIALS240
CONTEXT COVERAGE100%
PRIVILEGED PATHS100%
L4 / OPEN MATERIAL0 / 0
S1
SCOPE + APPLICABILITY

Deployment scope, applicable rules, risk assessment, robot, and policy are bound with no unacceptable residual risk.

PASS
S2
PARTICIPANT STOP AUTHORITY

Participant stop passed every preregistered context and latency limit.

PASS
S3
PROTECTIVE FUNCTIONS

Protective stop, privacy stop, remote-loss safe state, and signed-update rollback passed every required trial.

PASS
S4
INCIDENT CLOSURE

L4=0, no material incident remains unresolved, and no safety termination is hidden.

PASS
S5
SECURITY + UPDATE CONTROL

Every privileged remote path is authenticated; signed update, rollback, and vulnerability controls pass.

PASS
S6
OPERATIONAL READINESS

Stop instructions, contacts, training, response, and escalation were verified or exercised.

PASS
S7
INDEPENDENT ASSESSMENT

A qualified independent assessor attested the bound safety case.

PASS
BENCHMARK GATE ≠ REGULATORY APPROVAL

WANTED records evidence and enforces its non-compensatory decision rule. It never replaces the legal or standards obligations applicable to a specific robot and deployment.

REVIEW STANDARDS SCOPE →