AI Runtime Assurance

Start with one high-risk AI workflow.

Book a focused Agent Runtime Security & Evidence Sprint, then deploy runtime assurance where the risk is real.

The entry motion

Agent Runtime Security & Evidence Sprint

Start with one named workflow. Map the agent surface, identify runtime control gaps, prove the evidence path, then deploy runtime assurance where the risk is real.

1.0 0.5 0.0 policy threshold drift alert CUSUM · SLM v2.3 t · sampled inferences →

Workflow and drift readiness

The Sprint maps model calls, tool calls, drift signals, escalation paths, and operating events so your team knows where runtime assurance should be instrumented first.

req policy v2.4 permit deny escalate < 8ms · p99 CONTROL · v2.4

Runtime controls

Deployed in your environment. Local controls allow, block, redact, restrict, escalate, or require review before risky behavior reaches a workflow, tool, record, or customer.

FLEET · n = 12 1/12 drifting behaving drifting triggered

Security-review readiness

Focus the evidence on the buyer question: what the agent can do, what controls exist, what ran, what was blocked or escalated, and what remains to improve.

CUSTOMER PERIMETER ZERO SENSITIVE-DATA EGRESS your AI GLACIS witness observe · enforce prompts outputs user data stays in sha256 7f3e…d24b metadata content never crosses the wire

Zero sensitive-data egress

Glacis runs inside your infrastructure. Controls execute locally. Prompts, outputs, PHI, customer data, source code, credentials, and proprietary context stay inside your stack. Glacis exports only verification metadata, signatures, hashes, and evidence artifacts designed to prove control execution without exposing sensitive payloads.

OVERT-1.0 glc_7f3e… verdict: allowed prev: — 14:22:58 OVERT-1.0 glc_a1c0… verdict: allowed prev: 7f3e… 14:23:07 OVERT-1.0 glc_8e9f… verdict: flagged prev: a1c0… 14:23:21 OVERT-1.0 glc_____ pending… prev: 8e9f… 14:23:34 runtime audit-ready Evidence Pack

Evidence accumulates

Every control decision, drift signal, escalation, and signed receipt becomes review-ready evidence: evidence packs, security reviews, regulatory artifacts, and learning for the next control update.

Pricing built around the first risky workflow

Scoped engagement

Agent Runtime Security Assessment / Sprint

$48k fixed fee

A 10-business-day sprint on one AI workflow: agent/tool surface, runtime controls, evidence gaps, security-review readiness, and customer-ready proof.

Recommended
Deploy

Runtime Assurance Platform

From $60k/year

Local runtime controls, signed receipts, evidence packs, policy mapping, and zero sensitive-data egress.

Artifact

Evidence Packs

Custom / included by tier

Review-ready artifacts for enterprise security reviews, audits, regulated clinical AI evidence, insurance, and customer trust.

Founder-design-partner pricing is available for the first three customers — one named workflow, 10-business-day scope, 100% upfront, anonymized case-study and reference permission. Ask us on the call.

Not sure which tier fits? Grab 25 minutes with us — we’ll scope the workflow and evidence path.

Evidence receipts

Signed proof that runtime controls executed.

Runtime controls execute locally and generate signed receipts written to the OVERT 1.0 open standard. Receipts chain, tampering is detectable, and third parties can verify control execution without seeing the sensitive payload.

  • OVERT-conformant signed receipts, assembled into evidence packs
  • Framework exports: NIST AI RMF, ISO 42001, EU AI Act, OSCAL
  • Zero sensitive-data egress — evidence packs assembled from local receipts
Review evidence receipts

Pricing FAQ

What does the Sprint cover?

One named AI workflow over 10 business days: agent and tool surface, runtime controls, evidence gaps, security-review readiness, and a customer-ready proof artifact.

What happens after the Sprint?

Teams usually deploy runtime assurance where the risk is real: local runtime controls, signed receipts, evidence packs, and policy mapping for the workflow that needs enterprise trust.

We already use Vanta/Drata. Do we need GLACIS?

Vanta and Drata document that you have policies. GLACIS proves you followed them at runtime. They’re complementary — together they close the AI evidence gap.

Bring us one AI workflow.

We’ll map the agent surface, identify the runtime control gaps, and show what proof your customers will expect before they trust it.