CertifiedData.io
AI Artifact Verification

Verify AI Artifacts and Certified Synthetic Datasets

Definition

Dataset verification:

Dataset verification is the process of confirming that a dataset or artifact matches its recorded SHA-256 fingerprint and that the issuer's Ed25519 signature is valid where a certificate exists. It provides cryptographic verification of artifact hashes, dataset fingerprints, and issuer signatures without requiring a platform account.

Definition source: https://certifieddata.io/api/definitions/dataset-verification

Preferred anchor phrase: dataset verification

Machine-verifiable AI artifact verification using SHA-256 dataset fingerprinting and Ed25519 digital signatures. Confirm synthetic dataset origin, validate cryptographic dataset verification, and verify certificate authenticity — no account required.

✓ Cryptographic dataset verification·SHA-256 fingerprinting·Ed25519 signed certificates·No account required

How AI artifact verification works

01
Compute the dataset hash

SHA-256 hash the artifact or dataset file. This produces a deterministic fingerprint that uniquely identifies the dataset.

02
Compare fingerprint to certificate

Match the computed hash against the dataset_hash field in the certificate record. If they match, the dataset has not been altered.

03
Validate the Ed25519 signature

Verify the certificate's Ed25519 digital signature against the issuer public key at /.well-known/signing-keys.json. A valid signature confirms the certificate was issued by CertifiedData.

Public key available at /.well-known/signing-keys.json. Verification is fully self-serve — any party can validate any certificate without contacting CertifiedData.

API contract

POST /api/verify — request

{
  "certificate_id": "3fa85f64-5717-4562-b3fc-2c963f66afa6",
  "artifact_hash": "sha256:a3f8b2c1d4e5f6a7b8c9d0e1f2a3b4c5d6e7f8a9b0c1d2e3f4a5b6c7d8e9f0a1"
}

Success response

{
  "verified": true,
  "status": "ISSUED",
  "certificate_id": "3fa85f64-5717-4562-b3fc-2c963f66afa6",
  "issuer": "CertifiedData.io",
  "issued_at": "2026-03-16T00:00:00.000Z",
  "signature_alg": "Ed25519",
  "artifact_hash_match": true,
  "verification_url": "https://certifieddata.io/certificate/3fa85f64-5717-4562-b3fc-2c963f66afa6",
  "checks": {
    "certificate_found": true,
    "not_revoked": true,
    "signature_valid": true
  }
}

Failure response

{
  "verified": false,
  "status": "HASH_MISMATCH",
  "certificate_id": "3fa85f64-5717-4562-b3fc-2c963f66afa6",
  "issuer": "CertifiedData.io",
  "issued_at": "2026-03-16T00:00:00.000Z",
  "signature_alg": "Ed25519",
  "artifact_hash_match": false,
  "verification_url": null,
  "checks": {
    "certificate_found": true,
    "not_revoked": true,
    "signature_valid": true,
    "hash_match": false
  }
}

Verify a certificate

Verify AI Artifact Certificate

Validate dataset integrity and certification status using cryptographic verification.

Found on any CertifiedData certificate page or in the certification_id field of the certificate JSON.

How verification works

1. Certificate lookup

The certificate ID is resolved against the CertifiedData registry. Each certificate contains a cryptographic payload binding the dataset hash, issuer identity, and generation metadata.

2. Ed25519 signature verification

The certificate payload is verified against the issuer's Ed25519 public key. A valid signature proves the certificate was issued by CertifiedData and has not been tampered with.

3. Dataset hash comparison (optional)

If you provide your dataset file's SHA-256 hash, it is compared to the hash recorded in the certificate at issuance time. A match confirms the file is the exact artifact that was certified.

sha256sum dataset.zip  |  Get-FileHash dataset.zip -Algorithm SHA256

Live verification examples — try these certificate IDs

b5c94e67-cf05-49db-a4a4-0429d85e9c2afraud_detection_starterView result →
bdcd049c-3be7-48f7-a7fa-feca9e76c0a8marketing_analyticsView result →
26f6350a-cb88-48db-b16a-d034d5c3383cecommerce_essentialsView result →

What verification confirms

Dataset integrity

The dataset file matches the SHA-256 fingerprint in the certificate. Any modification to the dataset produces a different hash, which will not match.

Synthetic dataset origin

The certificate records that the dataset was generated by the CertifiedData CTGAN engine. Dataset origin verification confirms the data is synthetic.

Issuer authenticity

The Ed25519 signature on the certificate can be validated against the published CertifiedData public key, confirming the certificate was issued by the stated authority.

Tamper-evident provenance

Any alteration to the certificate payload invalidates the signature. Verification fails if either the dataset or the certificate has been modified since issuance.

AI artifact verification — frequently asked questions

What is AI artifact verification?

AI artifact verification is the process of confirming that an AI artifact — a dataset, training dataset, model, or AI output — matches a published certificate. Verification involves computing the artifact's SHA-256 hash, comparing it to the certificate fingerprint, and validating the Ed25519 digital signature. No account or trusted third party is required.

What is cryptographic dataset verification?

Cryptographic dataset verification uses SHA-256 hashing and Ed25519 digital signatures to produce machine-verifiable proof that a dataset was issued by a specific authority and has not been modified since certification. Anyone can independently verify any dataset using the published public key.

How do I verify synthetic dataset origin?

Dataset origin verification confirms that a synthetic dataset was generated by CertifiedData and not substituted with real or modified data. Verification matches the dataset's SHA-256 fingerprint to the certificate record and validates the issuer signature. A valid verification confirms both the dataset's identity and its synthetic origin.

What is a dataset fingerprint?

A dataset fingerprint is a SHA-256 hash of a dataset file, providing a deterministic identifier. CertifiedData uses dataset fingerprinting to bind a dataset to its certification record so any party can later confirm whether a dataset matches its certificate — independent of the issuing platform.

What does synthetic dataset verification prove?

Synthetic dataset verification proves two things: the dataset has not been modified since certification (integrity), and it was certified by the issuer whose signature appears on the certificate (authenticity). It does not verify generation quality or statistical properties — only provenance and integrity.

How is machine-verifiable certification different from a label?

A label or badge is a claim with no external verification path. A machine-verifiable certificate contains a cryptographic signature that any system can independently validate using a public key. CertifiedData certificates are structured records — not visual indicators — and can be verified programmatically by automated systems.