Every consequential control event can produce an OVERT-compatible receipt. Tamper-evident, independently verifiable, and designed so proof travels while sensitive data stays local.
receipt: { v: "overt/0.4", policy: "policy.fsi.credit_match", model: "llama-4-maverick@2026-04-01", decision: "allowed", at: "2026-04-25T10:51:57Z", prev: glc_887ab46a7f, sig: ed25519:9c3a…d04e }
Zero sensitive-data egress architecture
The receipt layer runs inside your environment alongside the runtime-control sidecar. It records control execution locally, generates a cryptographic commitment, and emits an OVERT-format receipt. Sensitive payloads stay inside your stack; only verification metadata, hashes, and signatures cross the trust boundary.
GLACIS is architecturally incapable of receiving your data
Open standard at overt.is
Crosses the trust boundary. Only hashes.
Independent verification. No vendor lock-in.
What runtime evidence provides
GLACIS generates signed runtime receipts in an OVERT-compatible format. Each receipt proves which control ran, what the outcome was, and when it happened—without exposing your data.
Open standard for AI attestation. Machine-readable, human-auditable, interoperable across tools. overt.is
Cryptographic commitments make any modification detectable. Receipts can’t be altered after the fact.
Any third party can verify a receipt without access to GLACIS. No vendor lock-in for your evidence trail.
Receipts roll up into reviewer-ready artifacts with control mappings, findings, remediation status, and verification metadata.
Your data stays in your environment. GLACIS emits only verification metadata, signatures, and cryptographic hashes across the trust boundary.
OSCAL-compatible evidence packs for auditors. Machine-readable, standards-based, ready for regulatory review.
Who this is for
AI vendors stuck in security review. Your product works. Their security team wants proof. Runtime receipts give them independently verifiable evidence that controls ran—not just a promise that they will.
Regulated organizations deploying AI. Colorado AI Act, EU AI Act, HIPAA—all require evidence of control execution. Runtime receipts preserve that evidence as a byproduct of operation.
Agent developers embedding governance. Your customers need proof. Hand them a verifiable evidence trail without building attestation infrastructure yourself.
Anyone whose AI decisions have consequences. If an AI output affects a person, a patient, or a financial outcome, you need signed proof that the right controls ran.
How you get it
The Agent Runtime Security Sprint instruments one high-risk workflow and turns signed receipts into a regulator-, auditor-, buyer-, and security-review-ready evidence pack.
Notarize receipts are part of every Enforce plan. See plans →
Open standard. Verify receipts without vendor lock-in. overt.is →
Frequently asked questions
An OVERT receipt is a structured attestation record in the open OVERT standard. It contains cryptographic commitments proving which controls ran, what the verdict was, and when—without exposing the underlying data.
Your data—prompts, responses, patient information—never leaves your environment. Runtime controls and receipt generation run locally. Only verification metadata, hashes, and signatures cross the trust boundary.
Anyone. OVERT receipts are independently verifiable. Your auditor, your customer, a regulator—they can verify without contacting GLACIS or using our tools.
An evidence pack is a reviewer-ready bundle of signed receipts, workflow context, control mappings, findings, and remediation status. It is built for security reviewers, auditors, buyers, regulators, and internal governance teams.
Receipts are most useful when they are attached to real runtime controls. The Sprint starts with one workflow so the evidence has an actual control surface, data boundary, and reviewer context.
Runtime assurance stack
Diagnostic
Identify attack paths, data boundaries, tool-call risks, control gaps, and the first receipts that need to exist.
View diagnosticControls
Drift detection, policy controls, permit/deny/escalate decisions on every request — each one witnessed and receipted.
Learn about controls