Evidence-Grade Enforcement
for Healthcare AI

We turn runtime controls into independently verifiable evidence that closes audits and procurement.

AI2 Incubator Cloudflare Launchpad Databricks

[email protected] • glacis.io

The Problem

AI Vendors Are Losing Deals to Procurement

Digital health vendors aren't getting blocked because their AI is unsafe. They're getting blocked because they can't prove it's safe in a way hospital security teams can verify.

The Vendor's Problem

Series A-C digital health companies losing deals, delaying deployments, burning cycles on manual audit responses.

The Buyer's Ask

"Prove your AI controls actually work—to someone who doesn't trust your dashboard."

The Gap

Guardrail vendors enforce and log. But the buyer still has to trust their interpretation. No third-party verification.

"Trust our dashboard"

What competitors offer

"Verify it yourself"

What GLACIS provides

Why Now

Regulatory Deadlines Create Urgency

June 30, 2026

Colorado AI Act

First US state law requiring "reasonable care" to prevent algorithmic discrimination. Healthcare AI deployers must demonstrate controls—not just claim them.

August 2, 2026

EU AI Act

Healthcare AI classified as "high-risk." Requires continuous automatic recording of events and traceability throughout lifecycle.

"If you could evaluate the top 10 AI tools I'm being pitched and give me one report that proves they're safe to deploy... sign me up."

— CMIO, 400-bed regional health system

The thesis: In 18 months, regulated AI without cryptographic proof will be uninsurable and unprocurable. We're building the infrastructure that makes deployment possible.

The Wedge

Evidence Pack Sprint: Days. Not Months.

Runtime Evidence Pack

Cryptographic proof that guardrails ran on every single request. Not logs—receipts.

Control Mapping

HIPAA, SOC 2, ISO 42001 pre-mapped. Auditor-ready from day one.

Buyer-Ready Portal

One URL replaces 400 security questions. Share it before buyers even ask.

Day 1

Drop-in sidecar

→
Days 2-8

Collect attestations

→
Day 10

Evidence portal live

Risk Reversal: No deployable evidence pack? Full refund.

We eat the risk because we know the tech works.

Inevitability

Trust Centers Become Proof Centers

2020

Trust Center

Static docs, SOC 2 badges, status pages

→
2025

Runtime Proof

Buyers demand evidence, not promises

→
2026

Proof Center

Table stakes for enterprise AI

The pattern is clear: Every B2B company has a trust page now. But those are promises about intent—not proof of execution.

We're not inventing a category.

We're riding an inevitable transition and owning the infrastructure layer.

How It Works

Zero-Payload Egress. No BAA Required.

We separate the Assertion (the AI output) from the Attestation (proof controls ran). Sensitive data stays in customer VPC. Only cryptographic commitments leave.

Customer VPC / Trust Boundary

Your App

GKE / ECS / Lambda

⟷

GLACIS SIDECAR

Attest → Hash → Sign

<50ms • 0 bytes PHI out

→

Witness

Hashes only

Merkle anchor

No BAA friction

PHI never leaves customer boundary. No Business Associate Agreement with GLACIS.

Third-party verifiable

Ed25519 signatures + Merkle roots. Security teams verify without trusting us.

Drop-in deployment

Sidecar alongside your gateway. Works with OpenAI, Anthropic, Vertex, any LLM.

Validation

Early Signal: Market Leader Reached Out

Credo AI's CEO (Forrester Wave Leader in AI Governance) reached out directly after seeing our positioning—meeting at JPM to discuss how our runtime attestation complements their governance platform.

When the market leader sees you as infrastructure they need, that's signal.

Design Partners

nVoq

Clinical voice scribe. Evidence pack pilot underway.

Prompt Opinion

AI governance for clinical decisions. Active design partner.

Pending

PraxisPro

AI governance for enterprise. Design partner.

Mayo Clinic

Platform CIO meeting scheduled.

Backing & Distribution

AI2 Incubator

$600K. AI-native company network.

Cloudflare Launchpad

GTM partnership. Infrastructure credibility.

Databricks

Credits + distribution. Healthcare AI focus.

CHAI Coalition

50+ health systems via Jennifer Shannon.

Expansion

Beachhead → Universal Infrastructure

Now

Healthcare AI

Hardest compliance = strongest moat

→
Next

Life Sciences

Pharma, med tech, clinical trials

→
Then

FinServ + Legal

SOX, GDPR are easier problems

→
End State

All Regulated AI

The proof layer for every AI decision

The SSL analogy

SSL became the trust layer for every web transaction. We're building the proof layer for every consequential AI decision.

Win the hardest wedge first. Healthcare AI → Enterprise buyers → Horizontal expansion. The playbook is proven.

Team

$250M+ in Combined Exits. Built the Last AI Keyboard 1B People Used.

Joe Braidwood

Joe Braidwood CEO

1 in 4 smartphone users globally

SwiftKey: Founding exec → $250M Microsoft acquisition

Cambridge Law • Regulatory navigation

Dr. Jennifer Shannon

Dr. Jennifer Shannon Co-Founder & Chief Medical Officer

First FDA-Authorized AI Diagnostic

Cognoa: Medical Director, secured FDA authorization

CHAI Coalition: Drafting AI governance standards

Caer Sanders

Caer Sanders Principal Engineer

Rust/WASM at Billion-Scale

WPI Robotics Faculty • Wayfair: Staff eng, <10ms ML

Advisors

Selvan Senthivel • GE Healthcare Chief Technologist
Nakis Urfi, JD, MPH • Cantex CCO (37 facilities)
David Márton • Harvard AI Research
Brett Murray • Nvidia, Walmart, Nokia C-Suite
John Ryley • Former Head, Sky News

Why this team wins: We've scaled consumer AI to 1B+ devices, navigated FDA authorization, and built <10ms production systems. We're not learning — we're executing.

Network Access

CHAI Coalition (50+ health systems) • AI2 portfolio intros • Cloudflare enterprise distribution

The Ask

$1.25M Angel Round

In 18 months, regulated AI without cryptographic proof will be uninsurable and unprocurable.

We're building the infrastructure that makes deployment possible.

[email protected] • glacis.io

1 / 10