For Risk / Insurance Teams
AI decisions become disputes. Verification reduces dispute cost and friction.
For defined high-impact lanes, Good Proof enforces externally verifiable, fail-closed gates so outcomes are defensible before they're challenged.
Verifiable decisions before claimsGuardian finality for high-impact exceptionsDecision-time Evidence PackScope-bounded control
Not a certification. Scope-limited verification. Acceptance depends on counterparty/programme requirements.
Risk 60-Second Brief
Problem
AI denials/exceptions can become disputes, appeals, and claim-handling cost.
Control
Status Link verification before reliance for defined high-impact action classes.
Output
Evidence Pack for claims, disputes, and oversight.
Boundary
Reduces friction/exposure; does not guarantee claim outcomes or model correctness.
Is This a Fit?
Best fit
- High-dispute action classes (denials, closures, payment exceptions, high-impact triage)
- Programmes requiring decision-time defensibility and stop-rely capability
Not fit
- Low-risk informational AI outputs
- Teams seeking model tuning without enforcement controls
Three Risk Outcomes
Reduced dispute friction
Verifiable decisions before they become claims.
Defensible exception handling
Guardian sign-off on high-impact exceptions with audit trail.
Bounded exposure
Fail-closed + immediate revocation reduces blast radius.
What goes into a dispute-ready file
Stamp ID + Status Link
action_class + scope boundaries
Current status + status-history references
signer, verified_at, expires_at, request_id, reason_code, failure_mode
Evidence Pack reference (IDA)
Minimal disclosure by default (no raw PII/PHI unless programme explicitly requires)
Status Triggers
NEEDS_REFRESH
Triggers
- Model or tool version update
- Policy/scope boundary change
- Key rotation or credential refresh
- Workflow version change (approvals, routing)
- External dependency version change
- Incident opened pending review
WITHDRAWN
Triggers
- Confirmed key compromise or integrity breach
- Scope boundary breach detected post-approval
- Material incident requires stop-rely
- Sanctions/allowlist breach discovered
- Critical vulnerability in verification stack
What to paste into RFPs and agreements
RFP Snippet (copy/paste)
"High-impact [ACTION_CLASS] SHALL require a valid externally-verifiable Status Link (No Stamp → No Ship)."
"On NOT_VERIFIED / NEEDS_REFRESH / WITHDRAWN / EXPIRED, Buyer SHALL treat as unverified and SHALL block or escalate per [ESCALATION_PATH]."
"Anti-spoof controls: verify_url host MUST match official_verifier; HTTPS/TLS required; redirects forbidden."
Webhooks accelerate orchestration; live Status Link verification remains the source of truth.
See /kill-switch for enforcement semantics and /clause-pack for exhibit templates.
Acceptance Criteria (Go-Live Pass/Fail)
Status Link resolves from official verifier
Fail-closed under timeout/TLS/domain mismatch
Non-VALID states block/escalate
Required evidence fields complete + exportable
Freshness + replay checks enforced (verified_at TTL / verify-per-action, request_id correlation)
If failed: Buyer MAY withhold reliance for affected action class.
Underwriting Evidence Readiness Checklist
Official verifier checks pass
Non-VALID fail-closed enforced
Evidence window and scope consistent
Exportable evidence fields
Dispute workflow and escalation defined
Stamp Sprint (30 Days)
One lane live: fail-closed gating + Status Link verification + dispute-ready Evidence Pack.
1
Scope highest-dispute action class + taxonomy2
Integrate verification, logging, evidence export3
Run fail-closed drills + refresh/withdrawal tests4
Deliver procurement packet + rollout planDefinition of done: Workflow blocks on status ≠ VALID for selected action class.
Risk reduced
- Disputes without decision-time evidence
- Unverified reliance on AI decisions
- Stale verification with no refresh trigger
- Exception handling with no auditable human trail
- Exposure from withdrawn/expired statuses
Out of scope
- Underlying model accuracy/correctness
- Regulatory/compliance certification
- Raw PII/PHI content validation
- Guaranteed claim outcomes/settlements
- Blanket certification/warranty
Good Proof is scope-limited verification. A VALID status confirms scope compliance under lane rules—not outcome correctness.
Assurance Governance (Optional by Lane)
Mind Chill Guardians
Programme-scoped human finality for the edge cases.
Programme-scoped human finality for exceptions/appeals only
Conflict checks + rotation
Multi-review thresholds for high-risk lanes
Minimal disclosure + auditable traceability
Ready to reduce dispute friction in high-impact AI lanes?
Definition of done: your workflow blocks on status ≠ VALID.
Scope-limited verification. Not a certification.