Service-Disabled Veteran-Owned

Control beneath
the mission-critical AI.

ALEETH Federal delivers the ICA Institutional Control Architecture to defense, intelligence, and federal missions, the layer that proves an AI-influenced decision was authorized, supervised, and defensible, on demand and after the fact.

Request a Briefing View Capability Statement

Cleared. Contained. Proven.

Times "trust the vendor's logs" has satisfied an accrediting authority

Mission AI does not get the benefit of the doubt. Every AI-influenced decision inside a defense, intelligence, or federal system already sits inside a named accreditation, supply-chain, or export-control obligation: CMMC, NIST 800-171, an ATO under RMF, DoDD 3000.09's autonomous-weapons review, ITAR.

The expectation is not future tense. It is the current duty to show, on demand, that an autonomous or AI-assisted action was authorized, within bounds, and reviewable. Not asserted in a policy binder, proven in a signed record.

The Architecture

Capture. Govern. Seal.

Cleared, contained, proven. ICA operates in three movements, and the same standard applies whether the decision is a contract review or a weapons-adjacent action.

01 · CAPTURE

Every consequential action

ICA sits beneath the AI system, not beside it. It sees the decision at the point it happens, not in a log shipped after the fact.

02 · GOVERN

Enforced, not logged

Policy is enforced in real time: allow, gate, halt, or quarantine. The control has the authority to stop a consequential action before it completes, not just record that it happened.

03 · SEAL

A record no one can rewrite

Every governed action is sealed into a signed, hash-chained record, independently verifiable and unalterable by the institution or by ALEETH.

Where It Maps

The federal & defense crosswalk

Six domains, each tied to a live accreditation, supply-chain, or export-control obligation.

Domain	Authority	What ICA proves
CMMC	DoD	Cybersecurity Maturity Model Certification compliance evidence across the AI-touched environment, not a point-in-time assessment.
NIST 800-171 / 800-172	Controlled Unclassified Information	Continuous control evidence over CUI-adjacent AI workflows, mapped to the required control families.
ATO / RMF Accreditation	Risk Management Framework	A signed control surface an Authorizing Official can examine directly. Accreditation evidence that doesn't go stale between assessments.
Responsible AI	DoDD 3000.09	Human-judgment and appropriate-levels-of-human-control evidence for autonomous and semi-autonomous systems.
Supply-Chain Integrity	SBOM	Provenance over the AI supply chain: what model, what version, what data, under whose authorization.
Export Control	ITAR / EAR	A data boundary and access trail proving controlled technical data did not cross an unauthorized AI surface.

Adversarial Test

We break our own agents first.

OPFOR is ALEETH Federal's adversarial range, owned by this company outright rather than licensed. Every agent put through it faces a standing adversary and is scored, and the result is written down whether it flatters us or not. An Authorizing Official gets a test record, not a vendor assurance.

Agents in the bone yard
Broken on the range

Agents under test

59.4%

Agent-scoped pass rate
303 of 510 trials

Critical escapes found

Escapes still open
Against one agent

Adversarial modules live

The bar, and why it is hard

Tier 1 is earned, never scored. An agent reaches it only by graduating, passing all nine gates, clearing 29,956 consecutive clean adversarial trials, and carrying zero critical escapes across its entire lifetime. A single critical escape is permanent and cannot be trained away. One agent holds Tier 1 today.

A separate battery of 90,000 trials certifies the shared policy gate rather than any individual agent. It is reported on its own line for that reason, and it is never counted toward an agent's record.

Measured on the live range, 14 August 2026. Published at opfor.aleeth.com.

Capability Statement

Who we are, in the format you need

ALEETH Federal, LLC

Institutional Control Architecture for Mission AI

Request Full Capability Statement (PDF)

Core Competencies

Real-time AI decision governance and enforcement
Signed, tamper-evident evidence chain (Ed25519, hash-chained)
Halt and quarantine of AI-influenced actions in production
Regulatory and accreditation crosswalk mapping
Agent and model registry across a governed environment

Differentiators

Control sits beneath the model, so it survives a model swap or prompt change
Evidence is independently verifiable, not a vendor-asserted log
Licensed from a live, in-production architecture, not a concept
Service-disabled veteran-owned and controlled, direct and unconditional

ENTITYALEETH Federal, LLC

STATE OF FORMATIONIowa

OWNERSHIP100% service-disabled veteran, direct

SDVOSB STATUSSBA VetCert application in process

IOWA TSBCertified 13 August 2026 (IEDA)

ICALicensed from The Schreck Group, Inc.

OPFOROwned by ALEETH Federal, LLC

UEI / CAGE CODEAvailable on request

NAICSAvailable on request

Iowa Economic Development Authority, Certified Targeted Small Business

Iowa Targeted Small Business

Certified by the Iowa Economic Development Authority on 13 August 2026. Iowa agencies may purchase directly from a certified TSB without competitive bidding, up to twenty-five thousand dollars per purchase, and sub-threshold opportunities reach us forty-eight hours ahead of the public bid site. Bid bonds are waived up to fifty thousand dollars.

Held by ALEETH Federal, LLC, an Iowa limited liability company. Valid through 13 August 2028.

Control beneaththe mission-critical AI.