RESOURCES — THE RECORD

Inspect the evidence. Read the limits.

Architecture, current product state, technical artifacts, and postmortems — published for scrutiny.

FORWARDABLE BRIEFS — NO FORM, NO EMAIL REQUIRED

One page each, written to be sent to someone who was not on the call. Print or save as PDF and the site chrome drops out.

Reading a decision record

What the record contains, how to verify a real one, and where the claim stops. For a contracting officer.

OT deployment & fail-closed behaviour

Where the Gateway sits, what happens on a DENY, and what only your pilot can establish. For an OT engineer.

AARM v1.0 alignment

Aligned, not certified: which requirements are satisfied, which is a gap, which is out of scope. For a compliance owner.

START HERE — THE RECORD, IN READING ORDER

01 WHITEPAPER — THE ARGUMENT → 02 TRUST — THE STATUS → 03 DOGFOOD — THE PROOF → 04 POSTMORTEMS — THE FAILURES

FOR YOUR SECURITY TEAM: TRUST + COMPLIANCE · FOR YOUR ENGINEERS: WHITEPAPER + BLOG · FOR YOUR CONTRACTING OFFICER: TRUST + POSTMORTEMS

DOC 01 · THE ARGUMENT

The full argument for deterministic runtime governance, and the architecture that delivers it.

DOC 02 · FIELD NOTES · ONGOING

Engineering notes on canonicalization, policy engines, and governing AI where the network ends.

DOC 03 · CANONICAL · REVIEWED QUARTERLY

Trust & Security

The canonical statement of our assurance status — what is claimed, what is in progress, what is not.

DOC 04 · FRAMEWORK MAPPING

Framework mapping and current posture across SOC 2, NIST SP 800-171, and the paths in progress.

DOC 05 · STANDING RECORD

How we run on Containment.ai

Our agent workforce runs under deterministic workflow controls. Action-guard enforcement is currently shadow/observe, and the gap is published.

DOC 06 · INCIDENT LOG

When something breaks, we publish what happened. A governance vendor that hides failures isn't one.

DOC 07 · SOLUTION · FRONTIER AGENTS

Frontier Agent Control

Sandboxes constrain where frontier agents run. MAG authorizes what they may make happen — deterministic, pre-execution control at the action seam.

MISSION BRIEFS & FIELD NOTE — UNGATED, NO FORM

BRIEF 01 · PDF · UNGATED

MAG — Frontier Agent Control

The product brief: deterministic authorization for frontier agents — what exists now, what is tailored per deployment, and the claim boundaries.

BRIEF 02 · PDF · UNGATED

MAG — Autonomy & Tactical Edge

The product brief for autonomy and tactical-edge missions: pre-execution authorization where the network ends and effects are physical.

FIELD NOTE · BLOG

Frontier agent control and the July 2026 incidents

What the OpenAI, Hugging Face, and Anthropic disclosures actually show — and where an authorization boundary fits.

THE EVIDENCE CENTER — ARTIFACT INDEX

The documents above are the argument. These are the artifacts — cryptographic material you can check yourself, in your own browser, without talking to us.

ARTIFACT	WHAT IT CAN ESTABLISH
Gateway receipt integrity package	One isolated page containing the local verifier, staging sample and tamper test, format documentation, and published staging key. It recomputes the receipt hash, verifies the Ed25519 signature against that key, and checks continuity of the supplied chain segment. It does not re-run policy, prove chain completeness beyond the supplied segment, or establish production deployment or authorization. Receipt checks run locally; the verifier makes no outbound data requests and never uploads receipt data. No account.
AARM Core self-attestation	Per-requirement conformance record against CSA's AARM v1.0 — including the gaps. The public attestation document is the controlling record.
Agent Governance Toolkit comparison	Where Microsoft's Agent Governance Toolkit and an in-path enforcement layer do different jobs — control-plane governance versus a deterministic ruling at the action seam.
Enforcement & audit evidence	The canonical assurance record — product-specific decision-record profiles, current availability, limitations, and the milestone status behind every claim on this site.

RECEIPTS, PER PRODUCT

MISSION AUTHORIZATION GATEWAY · FLAGSHIP

The Gateway staging build emits Ed25519-signed, per-organization hash-chained receipts at the edge seam. The browser verifier checks receipt integrity, the published staging key, and continuity of the supplied segment.

AGENT GOVERNANCE

The connected proxy issues HMAC-signed decision records, verified inside the deployment. Product scope and current status are stated on Trust.

AI CHAT FIREWALL

Every ruling writes an audit-ready, tamper-evident receipt — feeding the compliance program you already run.

The browser verifier and published staging key are Mission Authorization Gateway artifacts. It checks receipt integrity, signer key, and supplied-segment continuity; it does not re-run policy.

SEEN THE RECORD?

Now watch it rule on your workflow.

Request a 30-minute Boundary Review

Bring one AI workflow, one consequential action, or one data boundary.