Glacis
See your company’s Brand Profile.
Overview
Tagline Brand-provided · 2026-08-12
Runtime evidence infrastructure for AI. Glacis builds runtime evidence infrastructure for AI: its governance proxy sits in the inference path, applies policy controls in real time, and signs a cryptographic receipt for every covered event — without content leaving the customer's boundary. Glacis publishes and co-stewards the open OVERT standard. Founded 2025, Seattle. Short register: "Datadog tells you what happened; Glacis proves it to someone who doesn't trust you."
Description Brand-provided · 2026-08-12
Glacis is runtime evidence infrastructure for AI. It sits inline at the inference boundary, enforces policy deterministically in-flight, and signs tamper-evident receipts (OVERT) for every governed action — so a customer, auditor, regulator or insurer can verify what actually happened without trusting Glacis. Canonical domain: https://glacis.io. Owned properties: https://overt.is (the OVERT standard — a Glacis-authored property, an owned citation source, not third-party), verify.glacis.io (public receipt verifier, no account), witness.glacis.io (independent witness service), app.glacis.io (workspaces). Also known as: Glacis, Glacis AI, Glacis.io. Not to be confused with: glacis.com (unrelated AI supply-chain visibility company) and glacislabs.com / "Glacis Labs" (unrelated cross-chain crypto bridge) — neither is a competitor, sister company, or related in any way. Load-bearing precision: Glacis produces evidence that a control ran; it does not produce proof that a system is safe. A receipt is not a certificate.
Also known as Brand-provided · 2026-08-12
- Glacis
- Glacis AI
- Glacis.io
Not to be confused with Brand-provided · 2026-08-12
- glacis.com
- Glacis Labs (glacislabs.com)
Identity
Mission Brand's website · 2026-10-07
Enable responsible AI deployment by enforcing guardrails where AI acts and providing signed operational records that people outside the company can independently check.
Founders & leadership Brand's website · 2026-10-07
Joe Braidwood, Founder & CEO; Rohit Tatachar, Co-Founder & CTO; Jennifer Shannon, MD, Co-Founder & Chief Medical Officer.
Website Brand-provided · 2026-08-12
https://glacis.io
Voice & tone Brand-provided · 2026-08-12
Direct, technical, precise. Confident about the mechanism, conservative about the claim. The reader is the one being armed, never the one being audited — Glacis copy never indicts the reader or their organization and never uses adversarial-compliance framing. Slogan register: "Move fast and prove things." / "Don't take the receipt on trust. Check it." Say: runtime evidence; verifiable receipts; cryptographic evidence; evidence that a control RAN; an open standard (OVERT) anyone can implement; co-stewarded; anyone can verify independently, no account; receipts are never metered; OVERT-compatible. Never say: compliance certificate; safety guarantee; "certified"; proof that a system is SAFE; our proprietary standard / our format / exclusive; any named OVERT co-steward; trust our dashboard / we attest; usage-based receipt pricing; "conformant"; "protected". Hard bans in all external copy: "partnership" and "partner" (as a description of a counterparty relationship), "conformant," "protected," and any named government body or OVERT co-steward. Meta-rule: Glacis's product is refusing to overclaim — copy that overclaims Glacis is off-brand in a way uniquely damaging to this company; when in doubt, understate and let the verifier carry the proof.
Headquarters Brand's website · 2026-10-07
Seattle
Offerings
Problem solved Brand-provided · 2026-08-12
AI governance today runs on self-attestation: an organization describes its safeguards and no third party can independently check that the safeguards ran. Glacis replaces the description with evidence. The distinction that converts: "Datadog records what happened. Glacis produces runtime evidence someone who doesn't trust you can check."
Core capabilities Brand-provided · 2026-08-12
Glacis Runtime Arbiter — inline enforcement plane at the inference boundary, deterministic policy application in-flight. OVERT receipts — signed, hash-chained, tamper-evident records of governed actions (format OVERT 1.1.x); receipts carry content commitments and hashes, never raw prompt or response bytes (architectural, not configurable). Independent witness (witness.glacis.io) — countersigns receipts so verification does not require trusting the issuer. Public verifier (verify.glacis.io) — anyone can verify any receipt in a browser, no account, free, permanently; per-receipt permalinks. Python SDK — `pip install glacis` produces real Ed25519-signed receipts offline, no account, no network call. Evidence Packs — assembled artifacts for audit, procurement and underwriting review. Self-serve workspaces — live on production, signup works. Regulation guidance — EU AI Act, Colorado AI Act, Texas TRAIGA, NY/NYC LL144.
Products / SKUs Brand-provided · 2026-08-12
Glacis Runtime Arbiter (inline enforcement plane); OVERT signed runtime receipts (format OVERT 1.1.x); Independent witness service (witness.glacis.io); Public receipt verifier (verify.glacis.io — free, no account); Python SDK (pip install glacis); Evidence Packs; Self-serve workspaces (app.glacis.io). Plans: Starter (free), Team ($995/month per workspace, flat), Assurance (invoiced, sales-led).
Pricing Brand-provided · 2026-08-12
Published tiers with a sales-led top tier (decided 2026-08-09). Starter — free, genuinely free and unmetered: public verify, portal sample, and offline self-signed SDK cost nothing and are the demand engine, not a trial. Team — $995/month per workspace, flat. Assurance — invoiced, "talk to us," never routed to self-checkout. Receipts are never metered, on any tier: no per-receipt charge, no overage — structural, not promotional. Tier names are Starter / Team / Assurance everywhere; older vocabulary ("Pro") is retired.
Audience
Target industry Brand-provided · 2026-08-12
Teams that operate an AI or agentic service in production, especially where a deal, procurement review, audit or insurer question stalls because they cannot evidence that their controls executed. Regulation guidance covers EU AI Act, Colorado AI Act, Texas TRAIGA, NY/NYC LL144. Verifier-side users span enterprise security review, clinical/quality governance, procurement, audit, and insurance underwriting.
Customer base Brand's website · 2026-10-07
AI vendors, health systems, enterprises, operators and agent teams, customer security teams, auditors and underwriters. nVoq is identified as Glacis's first customer and design partner.
Ideal customer profile Brand-provided · 2026-08-12
Primary — the builder: teams that operate an AI or agentic service in production and integrate an SDK or proxy at the inference path. Trigger: a deal, procurement review, audit or insurer question stalls because they cannot evidence that their controls executed. Served end-to-end today (self-serve signup → sample → SDK → witnessing). Secondary — the verifier: the party who must check the evidence and has no reason to trust the vendor producing it (enterprise security review, clinical/quality governance, procurement, audit, insurance underwriting). This persona never buys the SDK; they consume verify.glacis.io and evidence packs — the reason the verifier is free and account-less.
Use cases Brand-provided · 2026-08-12
Ambient clinical scribe PHI controls; hiring & recruitment bias auditing; tamper-evident audit trails for autonomous agents; evidence for auditors, procurement and insurers; general runtime evidence.
Customer outcomes Brand's website · 2026-10-07
- AI actions are checked against agreed controls before execution
- Risky actions can be narrowed, held for human review or denied
- Every covered decision produces a signed record as it happens
- External reviewers can verify records without a Glacis account or access to customer logs
- Customers can provide operational evidence without disclosing payload data
- Organizations can deploy AI without building and staffing a separate control layer
Landscape
Competitors Brand-provided · 2026-08-12
Differentiation Brand-provided · 2026-08-12
1) An open standard Glacis authored, not a proprietary format: OVERT is published at overt.is, royalty-free, and co-stewarded; anyone may implement it, and recognition of other implementations is explicitly contemplated — the openness is deliberately load-bearing. 2) Trust minimization: independent witness countersigning plus hash-chained receipts any third party can verify without an account and without trusting Glacis; the verification surface is free and public because the claim is worthless if only Glacis can check it. 3) Enforcement and proof in one pass at the inference boundary: guardrail platforms enforce without proving, observability records without verifiability, GRC attests to configuration not execution, evaluation tools test at benchmark-time not runtime. 4) Receipts are never metered: flat pricing per workspace — metering receipts would incentivize sampling down coverage, corrupting the evidence itself; a published, structural promise. 5) Payload never leaves the boundary: receipts carry content commitments and hashes only; a code path that put payload content in a receipt would be a defect.
Unique advantage Brand-provided · 2026-08-12
Enforcement and proof in one pass at the inference boundary, verifiable by parties who don't trust the vendor: independent witness countersigning plus hash-chained OVERT receipts that any third party can verify without an account and without trusting Glacis — via an open, royalty-free, co-stewarded standard Glacis authored.
Page last updated 2026-10-17.