aiGEC™
The proof that your AI was governed. Per inference.
A cryptographic certificate for every governed output.
aiGEC™, the Governance Execution Certificate, is a machine-readable, cryptographically signed certificate generated as a structural output of the aiGUARD™ execution-control architecture.
Each certificate records the governance parameters applied to a specific AI output instance: the confidence score, the consequence classification, a hash of the user state, the execution decision, and the accessibility compliance result. The certificate is signed by a hardware-protected key inside the governance pipeline and is independently verifiable by any third party using the issuer's published verification key.
Possession of a valid, signature-verified aiGEC constitutes proof that the associated AI output was governed before delivery.
Three constraints. One guarantee.
A valid aiGEC™ cannot be generated by a system that has not evaluated the candidate output through the non-bypassable governance pipeline. This guarantee arises from three architectural constraints operating together:
Interface constraint
The certificate generation component exposes no external API or public endpoint. Its sole input interface is an authenticated internal message channel that accepts requests only from allowlisted callers within the governance pipeline. The response generation component is not on the allowlist.
Hardware key protection
The signing private key is generated within and stored exclusively in a hardware-protected key management facility (HSM or cloud HSM, FIPS 140-2 Level 3 or equivalent). The key never leaves the hardware boundary. Signing operations execute only when invoked by the certificate generation component's authenticated caller identity.
Output hash binding
Each certificate includes a SHA-256 hash of the delivered output. Any modification of the output after issuance renders the hash invalid, detectable by any party who recomputes the hash. The certificate is bound to one specific output instance and cannot be re-applied.
Verifiable by any third party. Without internal access.
aiGEC™ certificates are designed for decentralised verification. The issuer publishes the public verification key in JWKS format at a well-known HTTPS endpoint. Any third party, a regulator, an auditor, a downstream system, an enterprise procurement team, can independently confirm:
- → That the certificate was issued by the authorised governance pipeline
- → That the certificate fields are internally consistent and have not been modified
- → That the delivered output matches the hash recorded in the certificate
Verification requires no access to the issuer's internal systems. The certificate, the verification key, and the delivered output are sufficient.
Designed to fit the agent_attestation hook.
Mastercard and Google's Verifiable Intent specification (published February 2026) defines an optional agent_attestation field as an extension hook for carrying agent identity or security attestations. The field is currently undefined; content is deferred to future companion documents.
aiGEC™ is a fully-specified candidate for that field. An aiGEC carried as the agent attestation in a Verifiable Intent credential would give the payment network cryptographic proof that the AI output generating the transaction instruction was evaluated through a non-bypassable governance pipeline before delivery.
Integrate aiGEC into your governance pipeline.
aiGEC™ is available as a standalone licence and as part of aiGUARD™ deployments. Detailed technical specifications and integration patterns are provided under NDA.
Get in touch