Make every AI control independently verifiable.
Trust infrastructure for AI.
Built on ATLAS, the open standard for AI governance verification
GLACIS is the canonical implementation of ATLAS — an open standard for attestable AI governance. It turns controls into independently verifiable receipts. Start with an Evidence Pack for procurement and audit, then expand into continuous runtime verification as you scale.
“Your AI Needs an Alibi”
Washington State’s Chief Privacy Officer told a room full of healthcare builders what’s coming. A CMIO overseeing 12 hospitals asked us to build exactly what we’re building. A Medicare payer described our product without knowing it exists.
“We have to get certification from vendors that they’re using an AI governance program like NIST AI Risk Management Framework… In healthcare, you’re almost always walking into the high-risk space.”
KATY RUCKLE — Chief Privacy Officer, Washington State
Why verification matters
Independent verification is the missing layer
You already have policies, controls, and good intentions. GLACIS adds the structural property AI has been missing: third parties can verify that governance actually executed, without trusting the operator’s own logs.
From Policies to Proof
Your policies describe what should happen. GLACIS verifies what did happen — at runtime, on every AI interaction.
Independent, Not Self-Attested
Dashboards and logs are great for operations. GLACIS adds the independence layer — third-party witnessed evidence that stands up in regulatory and legal contexts.
Tamper-Proof Receipts That Controls Ran
GLACIS enforces controls at the inference and tool boundary and emits third-party witnessed receipts for every consequential decision. Like a flight data recorder — neither the operator nor the auditor has to trust your app logs.
The reasonable-care advantage
Colorado’s AI Act creates a rebuttable presumption of reasonable care for organizations that demonstrate NIST AI RMF or ISO 42001 compliance with verifiable evidence. The EU AI Act rewards the same posture. GLACIS gives you that evidence automatically — turning compliance from a cost center into a competitive advantage.
Better together
GRC Platforms + Runtime Proof
GRC platforms like Vanta and Drata document what you say you do. GLACIS adds independently verifiable, per-inference runtime receipts with third-party witnessing. Together, your compliance story is complete.
| Capability | GRC Platforms | GLACIS |
|---|---|---|
| Policy documentation | Documents what you say you do | Documents policies AND proves they executed at inference time |
| SOC 2 / ISO 27001 | Core strength | Maps to these + AI frameworks |
| AI-specific frameworks NIST AI RMF, ISO 42001 |
Limited or manual | ISO 42001 · NIST AI RMF · EU AI Act · Colorado AI Act — native mapping with automated evidence |
| Runtime evidence | First-party compliance evidence | Signed receipt per AI decision — model version, controls applied, policy outcome, timestamp |
| Real-time monitoring | Periodic evidence collection | Continuous attestation |
| Third-party witnessed proof | First-party evidence | Independently witnessed — per-inference runtime receipts with third-party signing, verifiable without trusting your app logs |
| Zero-egress architecture | Not applicable | Only hashes cross trust boundary |
| Colorado reasonable-care defense | Policies necessary, not sufficient | Evidence of NIST AI RMF adherence supports your showing of reasonable care |
| Independent verifiability | Policy documentation | Cryptographic evidence structured for audit and regulatory review |
| Customer data visibility | Platform-dependent | We never see your prompts, responses, or PHI — only cryptographic commitments. Zero-egress means plaintext content never leaves your environment. |
How it works
Every AI Decision, Proven
GLACIS evaluates AI requests against your governance policy — permit, deny, escalate, or flag. Each decision on configured controls generates independently witnessed evidence. Plaintext content never leaves your environment. Only cryptographic proof crosses the trust boundary.
You See Everything. We See Nothing.
What your team sees
- Full request and response content
- Policy evaluations with pass/fail details
- Complete audit trail, searchable by date, system, or outcome
- Dashboards for quality assurance and improvement
- All stored in your environment
What leaves your environment
- Only HMAC’d commitments (hashed, not reversible)
- Zero PHI. Zero request content. Zero response content.
- Cryptographic proof that controls executed — nothing more.
Evidence generation adds less than 50ms per inference. Storage costs less than your current observability stack.
See it in action
Watch Governance Enforced in Real Time
An AI request arrives...
How It Works
From policy to proof in three steps
Define your governance posture. GLACIS enforces it at runtime and seals every decision with a tamper-proof, third-party-verifiable receipt — evidence your auditors, regulators, and buyers can trust.
Zero Egress
Data stays local
Inline enforcement
Shadow to enforce
Tamper-proof
Crypto signatures
<50ms
Total overhead (p95)
What this unlocks
One verification layer. Three ways in.
Evidence Pack
Map controls, surface missing runtime evidence, and walk into procurement or audit with a defensible packet.
Learn moreContinuous Verification
Turn runtime receipts into continuous governance across ISO 42001, NIST AI RMF, EU AI Act, and buyer due diligence.
Learn moreZero-Egress Deployment
Embed once at the inference and tool boundary. The GLACIS sidecar attests AI decisions locally — only cryptographic commitments cross the trust boundary.
Learn moreWho we help
Verification for every stakeholder in your AI pipeline
Healthcare AI Vendors
Get through procurement faster with proof
Health Systems
Evidence for discovery — before you need it
Financial Services
SR 11-7 model risk with cryptographic proof
EU AI Act
Article 12 & 14 evidence, continuously proved
“We would never accept this for any other critical system. Financial systems have audit trails. Medical devices have mandated records. Aircraft have flight recorders. AI systems need the same level of verifiable evidence.”
As seen in
FAQ
Common questions
We already have SOC 2 / are working toward HITRUST
Great — those cover IT controls. AI-specific assurance addresses model behavior, decision audit trails, and content safety risks that SOC 2 and HITRUST don’t cover. They’re complementary.
How is this different from our existing documentation?
Documentation describes what should happen. GLACIS provides independent proof of what actually happened — verifier-ready receipts that your controls executed, not just that policies exist. Beyond evidence, GLACIS also enforces runtime controls — it doesn’t just prove what happened, it ensures the right thing happens in the first place.
What industries do you work with?
We work with AI teams in regulated industries including healthcare, financial services, insurance, and enterprise. The common thread is needing to prove AI controls work, not just that policies exist.
What if we’re not ready for a full attestation program?
That’s fine. We offer focused engagements for teams who need to unblock deals now. Start with what you need, expand later.
Does GLACIS just monitor, or does it actually enforce?
Both. You define controls declaratively — which guardrails to enforce, at what confidence thresholds, with what failure modes. The GLACIS arbiter evaluates every AI request against your active policy and makes real-time permit/deny decisions. Every enforcement decision is independently attested. You start in shadow mode (observe only) and transition to enforcement when ready. The transition itself is attested.
Proof, not promises
Colorado’s AI Act rewards evidence of reasonable care with a rebuttable presumption in your favor. At $20K per violation, $500/month is the easiest budget conversation you’ll have.
Evidence Pack
Show security, legal, and procurement exactly where proof exists and where it doesn’t.
See the Pack →Continuous Verification
Continuous receipts for every consequential interaction. Proof-of-Guardrail when reviews turn into investigations.
View Continuous Verification →Zero-Egress Deployment
Attest AI without moving data. Zero-egress sidecar, unlimited systems, dedicated support.
Learn More →