
When AI makes high-impact decisions, "trust us" isn't enough. Good Proof gives you contract-ready language and verification mechanisms, externally verifiable via the Official Verifier, without privileged access to vendor internal systems.
No Stamp → No Ship for defined high-impact lanes.
No certification claims · No access to vendor internals required · Programme-scoped · Buyer-controlled thresholds
Not a certification. Scope-limited verification. See the three acceptance carve-outs for the canonical scope statement.
Good Proof is not a blanket certification. It is scope-limited verification — process-evidence of what a Customer-defined Programme checked, not a warranty of underlying truth. Acceptance depends on counterparty, Programme, and contractual requirements. The three carve-outs below are the canonical scope statement and are designed to be referenced directly in evaluation memos and counterparty agreements.
Good Proof verifies what the Programme defines. The Stamp does not warrant that the underlying claim is fit for any particular use, merchantable, or otherwise suitable for a Reliance Decision the verifier did not author.
Good Proof emits Stamps as evidence of process completion against a Customer-defined Programme. Mind Chill disclaims, to the maximum extent permitted by applicable law, all implied warranties — including fitness for a particular purpose, merchantability, and results — with respect to the underlying subject of verification. Platform-level IP warranties and indemnities are addressed separately in the Master Verification Services Agreement.
A Stamp speaks only to the Programme-defined checks performed at the moment of issuance. It is not an ongoing assertion. Status changes, refresh triggers, and withdrawal travel via the live Status Link.
The Stamp attests to the result of running a Customer-defined rule set against submitted evidence at a specific point in time. It is not an opinion on facts outside Programme scope, not a continuing assertion of compliance with the Programme rule set, and not a forecast of the verified subject's future behaviour. Reliers must consult the live Status Link at decision time.
The party choosing to act on a Status Link makes the Reliance Decision. Good Proof provides evidentiary infrastructure — process, schema, signatures, status — not the judgment about whether to rely.
The Reliance Decision sits with the Relier. Good Proof furnishes evidence of process completion, scope, evidence-window, freshness, and revocation state. The decision to extend credit, bind cover, ship product, release funds, or otherwise act on a verified subject remains with the Relier and is governed by the Relier's own diligence, contractual posture, and applicable law, subject to mandatory consumer-protection rules where the Relier is a consumer.
For procurement and legal teams: These carve-outs reflect the contractual posture in Mind Chill's Master Verification Services Agreement (MVSA). The summary text is plain-language; the reference language is drafting-grade starter copy intended for adaptation by counsel. Final wording is subject to Mind Chill legal review.

In disputes: The Status Link is the "reliance state now." The IDA Evidence Pack is the "fileable snapshot for the decision-time record."
What you walk away with.
Exhibit-grade clause language with bracket variables
Status Links required in RFPs and agreements
Timestamped evidence exportable for disputes and oversight
High-impact [ACTION CLASS] SHALL require a valid Status Link (No Stamp → No Ship). • Fail-closed: if NOT_VERIFIED / NEEDS_REFRESH / WITHDRAWN / EXPIRED or verification fails → Buyer SHALL treat the decision as unverified and SHALL block or escalate per [ESCALATION PATH] • Anti-spoof: verify_url host MUST match official_verifier; HTTPS only; no redirects • Supplier must support refresh triggers on material change: [MODEL / POLICY / TOOL / SCOPE / KEYS / VERIFIER DOMAIN] • Evidence: IDA Evidence Pack for defined lanes [ACTION_CLASS_LANES] • Notification schedule: [NOTIFICATION_SCHEDULE] (Optional addendum: AI GOLD coverage may be layered for [DUTY_OF_CARE_LANES] per separate exhibit.)
Complete bracket variables in your Order Form. See /clause-pack for full exhibit templates.
Status is returned wherever the Status Link is checked. Enforcement latency is bounded by verify-per-action/TTL policy.
"AI Verification & Kill-Switch Requirements"
Bracket variables (e.g., [ACTION CLASS], [NOTIFICATION_SCHEDULE]) are completed in the Order Form. See /clause-pack for negotiation-ready clause templates.
Enforcement semantics (fail-closed behavior, TTL, webhook propagation) are defined at /kill-switch.
Pass/fail at go-live
If criteria fail, Buyer may withhold reliance for the affected action class.
Minimum fields per decision
Designed for underwriting, disputes, and oversight. Verification is scope-limited; no PII/PHI required by default.

One lane live: fail-closed gating + Status Link verification + IDA Evidence Pack.
Define 3 action classes, contract exhibit variables, success criteria
Wire verification checks, logging, evidence exports
Timeout/TLS/domain mismatch tests, refresh/withdrawal drills, tabletop
Exhibit text, evidence specimen, rollout plan
If you send us your standard AI vendor addendum, we'll return a redline-ready Exhibit that drops in cleanly.
Good Proof is scope-limited verification. A VALID status confirms scope compliance under lane rules—not outcome correctness.
Expand to review optional governance mechanisms
These governance mechanisms are optional and enabled per lane based on programme requirements.


Programme-authorised human finality for the edge cases—only when required.
30-day Sprint. One lane live. Contract exhibit + evidence specimen + rollout plan.
Not a certification. Scope-limited verification.