Glacis Technologies, Inc.
Glacis Technologies provides runtime supervision and verifiable evidence for AI systems by applying approved rules at the point of action, allowing, blocking, redacting, escalating, or requiring human review, and creating signed operational records that can be independently verified. Its platform enables operators, vendors, and reviewers such as customers, auditors, regulators, and insurers to check how AI systems were controlled without relying on proprietary logs or claims. Glacis aims to deliver assurance and transparency for high-stakes AI deployments through open standards and independently checkable records.
Facts
- HQ
- Seattle, WA, US
- Sectors
- Artificial Intelligence
- Founded
- 2025
- Employees
- 8
- Stage
- startup
- Founders
- Joe Braidwood, Rohit Tatachar, Jennifer Shannon
Is this your organization? Claim it
Joe Braidwood announces Glacis built on Cloudflare Workers to enable low-latency supervision and signed operational records
Joe Braidwood explains why Cloudflare Workers powers Glacis, describing how the distributed edge enables low-latency supervision, signed operational records, and built-in managed security controls.
AI Supply Chain Security: Guide Explains Risks Across Data, Models, Libraries, Infrastructure, and Tooling
This guide explains AI supply chain security, extending software supply-chain practices to AI-specific components like datasets, pre-trained models, libraries, infrastructure, and tooling, and highlights attack vectors such as data/model poisoning, dependency confusion, and provenance attacks.
EU adopts AI Omnibus; Annex III high-risk rules apply 2 Dec 2027 and Annex I product rules 2 Aug 2028
The EU AI Act (Regulation 2024/1689) is in force with phased application: prohibited practices effective from Feb 2025, GPAI model obligations from Aug 2025, and the AI Omnibus (Regulation 2026/1744) set Annex III high-risk obligations to apply from 2 Dec 2027 and Annex I product-embedded obligations from 2 Aug 2028; organizations should classify systems and map applicable provisions rather than treating those dates as a single deadline. By August 2026 the GPAI Code of Practice and its public register are live, xAI appears on the register for one chapter, and harmonised standards are expected in Q4 2026.
Colorado signs SB 26-189 replacing 2024 AI Act with ADMT disclosure regime; obligations start Jan 1, 2027
SB 26-189, Colorado’s 2026 Automated Decision-Making Technology law, repeals the 2024 Colorado AI Act and institutes a narrower transparency- and disclosure-focused regime for developers and deployers of ADMT, with substantive obligations beginning January 1, 2027. It replaces the 2024 law’s high-risk/reasonable-care model with duties on notice, documentation, and consumer rights and reflects legislative pushback against a burdensome impact-assessment approach.
Joe Braidwood warns ambient AI scribes create malpractice risk as hallucinations and encounter-level outputs alter medical documentation
Joe Braidwood examines how ambient AI medical scribes can streamline clinician workflow but introduce documentation and medicolegal risks, since encounter-level AI outputs affect what can be reconstructed later.
Joe Braidwood: SOC 2 and HITRUST Don’t Cover AI Risks — Compliance Leaves Critical Gaps
Joe Braidwood explains that SOC 2 and HITRUST, while important for IT security, were not created to address AI-specific risks, leaving a gap in protections organizations must recognize and manage.
Joe Braidwood explains using signed, scoped operational records to tie AI policy to control decisions without certifying compliance
Joe Braidwood explains how signed, scoped records can link AI policies to specific control decisions, clarifying evidentiary practices without claiming those records certify safety, effectiveness, or compliance.
Joe Braidwood on GLACIS Proposes Three-Layer AI Security Framework: Prevention, Visibility, Evidence
Joe Braidwood outlines a three-layer AI security framework—prevention, visibility, and independently verifiable operational evidence—and specifies the claim boundaries each layer requires.
Joe Braidwood details how inline redaction, de-identification review, and HIPAA-ready vendor configurations can limit BAA exposure in healthcare AI
Joe Braidwood outlines practical measures—inline redaction, de-identification review, and HIPAA-ready vendor settings—to lower risks when sharing healthcare data with AI tools, while cautioning against overclaiming legal protections.
People
-
Joe Braidwood Founder & CEO
-
Jennifer Shannon Co-Founder & Chief Medical Officer