Verifiable AI evidence
Trust Standard Protocol gives AI systems a standard way to produce tamper-evident, portable evidence of what happened — so teams, auditors, customers, and regulators can verify the record without trusting the vendor.
Verified — the seal holds Failed — one byte changed, the seal broke
offline · real Ed25519
AI governance is moving from policy statements to evidence. TSP gives teams a standard way to preserve what happened, bind it cryptographically, and make it independently checkable later.
For teams working with the EU AI Act, DORA, internal controls, procurement reviews, and incident response, the question is no longer only whether a system performs. It is whether the record can be examined, challenged, retained, and trusted after the fact.
The durable advantage is not only capability. It is provability.
The problem · The fix
Logs, dashboards, and internal audit trails may be useful operationally, but they are difficult for an outside party to verify independently once a dispute, review, or regulatory question appears.
TSP turns AI-mediated events into tamper-evident receipts that can be checked independently and offline — giving teams a portable evidence layer for auditability, oversight, investigations, and trust.
If an AI system made or supported a decision, the record should be independently verifiable.
How it works
A TrustEnvelope records the event in canonical form and seals it cryptographically so later changes are detectable.
Receipts can be chained and, where needed, independently anchored to strengthen evidentiary posture beyond a self-attested local record.
The cryptography can be re-run offline, producing the same integrity result for every verifier.
What you get
If the signed record changes, verification fails. Quiet edits do not survive inspection.
Evidence can move across teams, customers, auditors, and regulators without depending on the issuing vendor's dashboard.
Verification does not require trust in us or in the issuer — only the published rules, keys, and cryptography.
Why TSP
Each event is sealed as a TrustEnvelope — canonical JSON, hashed with SHA-256, signed at the point of issuance.
Anyone can re-run the Ed25519 verification offline. The evidence is not trapped behind an API, account, or vendor workflow.
The trail stays legible to the people who must assess it — engineering, compliance, audit, procurement, and oversight.
Official status is governed separately from technical verification. Payment never grants it, and verification alone does not confer it.
The systems-level view
Every AI deployment is moving toward one of these models. Only one produces portable, independently verifiable records.
Outputs matter, but the evidence trail is incomplete, fragile, or absent.
Each organization builds its own evidence workflow, but verification remains local and non-portable.
Outputs carry portable, cryptographically verifiable evidence that different parties can inspect independently.
The honest boundary
See it for yourself
This verifier runs entirely in your browser against the issuer's Ed25519 public key. Nothing is uploaded. Change one byte and the verification result fails.
Real Ed25519 + SHA-256 · offline verification · no hidden trust in the issuer
The honest line
TSP does not replace legal, governance, or compliance programs. It gives those functions something they often lack: a portable technical evidence layer at the point where outputs are created and reviewed.
And one boundary keeps the system defensible: verification is not truth. A receipt can prove what a system recorded, signed, and declared at a moment in time — not whether the underlying decision was substantively correct or legally justified.
Narrow honesty
Useful systems define their boundary precisely. TSP covers the technical evidence layer; it does not claim to perform the legal, organisational, or adjudicative parts around it.
Covers — the evidence layer
Doesn't — by design
Who uses it
Add a verifiable evidence layer to AI workflows without making verification depend on your own platform.
Work from evidence that can be retained, reviewed, and challenged with a clear technical boundary.
Inspect a decision trail independently, without relying on screenshots, vendor claims, or internal dashboard access.
Exchange AI records across organizational boundaries with a shared verification method.
Make AI evidence verifiable before a regulator, customer, auditor, or counterparty asks for it. The reference verifier is open and runs offline.