YIN-AI CERTIFICATION
Cryptographic proof of AI authorization.
A signed certificate for every decision your AI makes. Issued in seconds. Verifiable by anyone, offline, forever, with no dependency on us.
Your email is used only to send your API key. We never share or sell it.
{
"certificate_id": "cert_9f2a4c…c07d",
"schema_version": "1.4",
"decision": "ALLOW",
"acr_score": 92.4,
"issued_at": "2026-07-28T14:07:19Z",
"authorizer_identity": {
"credential_type": "ORG_PKI",
"role": "Case Officer"
},
"authorization_scope": {
"process_type": "Housing Benefit Processing",
"regulatory_framework": "EU AI Act Art. 14"
},
"authorization_validity": {
"valid_until": "2026-07-28T18:00:00Z",
"single_use": true
},
"compliance": [
"EU AI Act Art. 9, 12, 13, 14, 50, 72, 86",
"NIST AI RMF",
"SEC Rule 17a-4",
"HIPAA",
"ISO IEC 42001",
"DORA",
"NYC AEDT 20-871"
],
"ed25519_signature": "d8266364…2622"
}
The same certificate, in plain English
- A named case officer authorized it. Confirmed with their organization's own PKI credential, not a username and password.
- Only for housing benefit processing. The scope was declared before the AI acted, under EU AI Act Article 14.
- Valid until 18:00 that day, and usable once. After that the authorization expires on its own.
- The AI was allowed to act, rated 92.4 out of 100. Change one character of this record and the signature stops matching.
Join the first organizations certifying AI decisions
For business and compliance teams
Your AI makes decisions every day. When a regulator, an auditor, or a customer asks whether a person authorized one of them, a log file is not proof. It can be edited, and it asks everyone to simply trust you.
- What the certificate proves. That the decision was authorized by a person, at the moment it happened, in a form that cannot be changed afterward.
- Who accepts it. Regulators, auditors, and courts. It stands on a published public standard, so nothing rests on trusting us.
- Why it matters now. New rules require evidence of human oversight for high risk AI. This is that evidence, as one signed document you can hand over.
For developers
- 1Send a request
POST your decision with your API key.
- 2Evaluation
It runs through the evaluation pipeline.
- 3Certificate returned
A signed certificate comes back in the response.
- 4Verify offline
Check it anywhere with the free yin-cert-verify tool. No call home.
- EU AI ActArt. 9, 12, 13, 14, 50, 72, 86
- NIST AI RMFUS risk framework
- SEC Rule 17a-4Records retention
- HIPAAHealth data
- ISO/IEC 42001AI management
- DORAEU financial resilience
- NYC AEDT 20-871Automated hiring
Nine modules. One certificate.
- 01Answer and Reasoning
Certifies that a person authorized your AI to answer, and records the reasoning it gave. Use it wherever a decision has to be defensible to a regulator, an auditor, or the person it affected.
- 02Vision AI
Certifies what a vision model saw and how confident it was, with a statistical coverage guarantee rather than a bare score. Use it where a misread image has consequences, such as medical imaging, quality inspection, or document processing.
- 03Predictive and Time Series AI
Certifies a forecast together with the drift checks that were run against it. Use it when a model's accuracy decays over time and you need proof it was still fit for purpose when the call was made.
- 04Tracking and Identification AI
Certifies who or what was identified, with a tamper-evident chain of custody. Use it where identification is legally sensitive, with raw identifiers hashed to meet GDPR Article 9 and biometric privacy law.
- 05Industrial and Manufacturing AI
Certifies that a control decision stayed inside a declared physical safety envelope. Use it on plant and process equipment, where an out-of-bounds instruction damages hardware or injures people.
- 06Mobility and Autonomous Systems
Certifies that the system was operating inside its declared design domain, and records the residual risk that was accepted. Use it for driver assistance and autonomous driving, where the question after an incident is what the system was authorized to do.
- 07Financial and Trading AI
Certifies a financial decision against model risk and market conduct requirements. Use it for credit, underwriting and trading, where SR 11-7 and MiFID II expect documented oversight and timing evidence.
- 08Defense and Military AI
Certifies a mission authorization as a chain, with a separate token issued at every level of command. Use it where authority has to be provable after the fact and no single actor can claim more than they were given.
- 09Smart Cities and Public Infrastructure
Certifies a decision that depends on other systems, carrying confidence across the whole declared chain of dependencies. Use it for utilities, transport and public services, where one model's output becomes another model's input.
Connect via MCP
YIN-AI CERTIFICATION is available as a Model Context Protocol server. Connect Claude, Cursor, or any MCP client directly to our certification platform with three tools:
yin_cert_certify— issue a certificateyin_cert_verify— verify offline, no key neededyin_cert_get_schema— fetch module schema
MCP endpoint: https://certification.yintechnologies.com/mcp
How it works
- Step 1
Get your API key.
Enter your email. Your key arrives in seconds. No credit card.
- Step 2
Add one API call to your application.
Your system sends the AI output to our platform automatically. No manual steps. No human in the loop for routine decisions.
- Step 3
A signed certificate comes back instantly.
It proves who authorized the AI to act, within what declared scope, and when. It stays in your systems.
- Step 4
Anyone can verify it offline, forever.
Your auditor, your regulator, your customer. No call home. Nothing to trust but the signature.
One certification stack, three layers
A signed decision is the first question an auditor asks, not the last. The same certification method extends upward: from the individual AI decision, to your organization's compliance posture, to whether that posture survives procurement review. Each layer consumes the one below it.
Did this AI decision carry valid authorization before it executed?
Available nowAre your compliance claims backed by verified evidence, or by self-attestation?
In developmentWill that compliance state survive federal procurement or adversarial review?
In developmentOnly the answer layer is available today. The compliance and
procurement layers are in development, and nothing you certify now depends on them —
an ACR certificate is complete on its own terms. The published chain is readable at
/v1/compose.
Download the offline verifier
yin-cert-verify checks a certificate's signatures on your own machine. No account, no API key, no network call. It rebuilds the exact signing preimage and verifies both the Ed25519 and the ML-DSA-87 signature against the issuer key.
curl -sSLO https://certification.yintechnologies.com/download/yin-cert-verify-linux-x86-64 curl -sSLO https://certification.yintechnologies.com/download/SHA256SUMS sha256sum --ignore-missing -c SHA256SUMS # expect: yin-cert-verify-linux-x86-64: OK chmod +x yin-cert-verify-linux-x86-64 ./yin-cert-verify-linux-x86-64 --json certificate.json
Exit code 0 means the signature is valid. Swap the filename for yin-cert-verify-macos-x86-64 or yin-cert-verify-macos-aarch64 on macOS.
Why regulators and auditors accept it
Every certificate is issued against the Authority Confidence Rating framework, a published standard with a permanent citable reference at Zenodo DOI 10.5281/zenodo.19361421.
Any regulator, auditor, or court can read the full specification independently. The certificate references the standard. The standard is public. The signature is verifiable offline. Nothing depends on trusting us.
Built by Ilyes Mazari, inventor of the Y.I.N. Mazari Architecture. 41 USPTO filings. 4,514 claims. Priority November 23, 2025.
What it does
Authorization confidence. Proved. Not logged. Not documented. Proved.
Free tier
100 signed certificates every month. No credit card.
Your email is used only to send your API key. We never share or sell it.
Enterprise
Unlimited volume, a custom service level agreement, and on-premises deployment where decision content cannot leave your network.
Framework
Authority Confidence Rating. Zenodo DOI 10.5281/zenodo.19361421.