AI Assurance You Can Prove
Security teams don't trust policy documents. They trust evidence. We help AI teams generate proof their controls actually ran.
What We Do
Proof, not just documentation
We generate verifiable proof that safety controls executed — timestamped, cryptographically signed, and mapped to the frameworks buyers recognize.
Not logs. Receipts. Cryptographically signed, independently verifiable. When the auditor or attorney asks "prove it"—you can.
The Problem
When things go wrong, no one can prove what happened
~100,000 patients. Allegation: ambient scribe hallucinated consent. Sharp can't cryptographically prove otherwise.
Major health system's AI Governance Committee blocked scribe vendor deal over data ownership concerns.
The most safety-conscious AI vendors face the same exposure. When the lawsuit comes, good intentions don't hold up. Cryptographic proof does.
Why Now
Regulatory deadlines create urgency
Every AI interaction touching PHI needs auditable proof
Affirmative defense for NIST AI RMF. $20K/violation
Healthcare AI = "high-risk." Continuous traceability required
ADMT opt-out. Risk assessments for healthcare decisions
The thesis: In 18 months, regulated AI without cryptographic proof will be uninsurable and unprocurable.
Standard of care precedes law. NIST AI RMF and ISO 42001 are already the de facto duty of care.
The Platform
Production-Ready. Shipping Now.
Live at app.glacis.io — governance platform, attestation feed, and witness network operational.
The SSL Analogy
SSL certs prove identity without dictating what you serve. GLACIS receipts prove enforcement of policies and guardrails at inference—without dictating your models.
Positioning
Platform, Not Product
Build PHI detection models
Compete with Comprehend/Macie
Own the guardrail logic
Attest models and guardrails
Ship strong defaults (open source)
Prove execution, not policy
Comprehend, Macie, open source models—we attest them all. At scale, we become the neutral trust layer for AI governance across the ecosystem.
The SSL Analogy
SSL certs prove identity without dictating what you serve. GLACIS receipts prove enforcement of policies and guardrails at inference—without dictating your models.
Traction
Design Partners in Healthcare AI
Colorado-based ambient scribe for home care • 50k+ visits/month
PraxisPro
In diligence
Prompt Opinion
In diligence
All from HLTH 2025 • Healthcare AI vendors facing procurement friction
Colorado AI Act: June 2026
Compliance pressure creates urgency
70+
Patent claims filed
4 families • Fenwick & West
Defensibility
Building the Category Definition
Co-Epoch Attestation
Binds receipts to specific binary versions. Prevents "binary substitution" attacks.
Statistical Safety Signals
Mathematically rigorous sampling for continuous compliance verification.
Zero-Payload Egress
Cryptographic commitments without data exposure. PHI never leaves boundary.
Federated Witness Network
Distributed attestation for independent verification at scale.
Filed Nov 2025 — First-mover advantage in attestation-based AI compliance.
Team
FDA Authorized. Enterprise Deployed.
We've Lived This Problem.
Joe Braidwood
Co-Founder & CEO
1 in 4 smartphone users
SwiftKey founding exec → $250M Microsoft • Cambridge Law
Dr. Jennifer Shannon
Co-Founder & Chief Medical Officer
FDA De Novo authorized
Cognoa: SaMD reimbursed
Caer
Rust & Cryptography Lead
10+ years production • WPI
Atreya
AI Engineer
Model eval & red teaming • WPI • Ex-PAAI
Advisors
The Ask
Pre-Seed: $1.5M
3
Design Partners
70+
Patent Claims
18mo
First-Mover Window
USE OF FUNDS
→ Close nVoq to paid contract
→ Ship cloud-native sidecar
→ Build compliance evidence packs
→ 10 vendors by Q3 2025
EU AI Act
Aug 2026
Colorado AI Act
Jun 2026
12-18 months to implement. The window is closing.
AI Assurance You Can Prove