The AUTHREX Application Set

One framework.
Five worked examples.

AUTHREX is one governance framework, a single verified authority pipeline. It ships as one product, AUTHREX-AGENT, and is applied to five documented U.S. federal needs as five worked-example applications, so this page presents the product plus five applications, six sections in all. They are not six products; they are one idea shown six ways, each anchored to a real DOI-registered platform and a simulation you can run. Lead with two: ASSURE and ICS-GATE carry the strongest, deadline-backed national-importance cases. The other three applications show the framework's reach.

Lead Applications

The two strongest cases.

These two have the clearest, most defensible national-importance arguments: a federal gap named in primary-source guidance, a statutory deadline, and a clean distinction from everything else on the market.

Supporting Applications

Three more, showing the reach.

The same pipeline applied to three further domains. Each is real and runnable; together they demonstrate that one governance pattern generalizes across very different problems.

In Full, With Diagrams

Each application, end to end.

The product and the five applications below, each with the diagram that frames how its authority decision works. Open any page to run its live simulation.

Software Governance Companion · The Product

AUTHREX-AGENT

Authority lifecycle governance for agentic AI.

The same seven-stage pipeline that governs autonomous aircraft and self-driving vehicles, instantiated as a software shim around LLM-based agents. AUTHREX-AGENT wraps any agentic AI runtime with the SATA → ADARA → IFF → HMAA → MAIVA → FLAME → CARA pipeline. No FPGA, no SBC, no model retraining required.

Reference architecture aligned with CISA + NSA + Five Eyes joint guidance Careful Adoption of Agentic AI Services (1 May 2026). This is the product; the five applications below are this same pipeline pointed at documented federal needs.

7
Pipeline Stages
4
Authority Tiers
15
Spec Sections
0
Hardware Deps
◇ THE PIPELINE
AGENT ACTION SATAInput trust scalar τ ∈ [0,1] ADARAPrompt-injection detection IFFTool authentication HMAATiered authority T3→T0 MAIVASub-agent quorum gate FLAMEDeliberation window CARARecovery on anomaly ERAM · CROSS-CUT EXECUTEall passed DELAYFLAME hold HANDOFFto human ABORTCARA recovers ECDSA-P256 SIGNED · APPEND-ONLY LEDGEREVERY DECISION COMMITTED BEFORE EXECUTION
CISA · NSA · 5-EYES
Careful Adoption guidance
NIST AI RMF
+ NDAA §1513 + DoD RAI
Federal-Need Application · Pre-Deployment Assurance

AUTHREX-ASSURE

Pre-deployment authority governance for autonomous systems.

The same seven-stage pipeline that governs autonomous aircraft and agentic AI, applied to the single most consequential moment in an autonomous system's life: the transition from test into production. AUTHREX-ASSURE gates that transition behind an explicit, signed assurance decision, recorded to the ERAM ledger, so no autonomous system reaches a live environment without a documented authority-to-operate.

Reference architecture aligned with the 2026 National Cybersecurity Strategy, which explicitly names the pre-deployment validation gap, and NDAA §1533, which directs a standardized DoD AI assessment framework due June 2027.

7
Pipeline Stages
4
Authority Tiers
T→P
Test-to-Prod Gate
2027
NDAA §1533 Due
◇ THE ASSURANCE GATE
CANDIDATE SYSTEMin test environment SEVEN-STAGE ASSURANCE CHECK SATA · input provenance attestedADARA · adversarial robustness screenedIFF · tool / interface identity verifiedHMAA · authority envelope boundedMAIVA · consensus integrity confirmedFLAME · deliberation window enforcedCARA · rollback path pre-armed ALL GATES PASS? YESNO SIGNED CERTIFICATEECDSA-P256 · ERAMauthority-to-operate HELD IN TESTno production pathfindings logged CLEARED · PROD EVERY DECISION COMMITTED TO THE SIGNED LEDGERBEFORE ANY PRODUCTION RELEASE
2026 NCS
Names the validation gap
NDAA §1533
Framework due June 2027
◇ THE IT/OT AUTHORITY BOUNDARY
IT ZONEAI proposes anOT action OTlive grid /water SCADA ICS-GATEAUTHORITY BOUNDARY if authorized SATAprovenance verified HMAAtier by criticality FLAMEhuman window CARArollback pre-armed ACTION REACHESCONTROLLER REFUSED ATBOUNDARYspoofed reading oruntrusted action AUTHORIZES THE SAFETY LOGICnever makes the safety decision itself CISA / NSA AI-IN-OT PRINCIPLES · 3 DEC 2025NIST SP 800-82 · ISA/IEC 62443 · NERC CIP
CISA · NSA
AI-in-OT principles
SIL 3 / NERC CIP
BLADE-INFRA-OT
Federal-Need Application · Critical Infrastructure

AUTHREX-ICS-GATE

Operational-technology authority governance for critical infrastructure.

The same pipeline that governs agentic AI, instantiated at the IT/OT boundary. When an AI system proposes an action on a live power-grid or water-treatment controller, AUTHREX-ICS-GATE decides whether that action is authorized at the current tier, with what evidence, and with a pre-armed rollback, before it reaches the controller. The cleanest distinction in the package: it authorizes the deterministic safety logic, it never makes the safety decision itself.

Reference architecture aligned with the CISA / NSA "Principles for Secure Integration of AI in Operational Technology" (3 Dec 2025). Cross-walks to NIST SP 800-82, ISA/IEC 62443, and NERC CIP.

7
Pipeline Stages
4
Authority Tiers
IT/OT
Boundary Gate
SIL 3
Platform Class
Federal-Need Application · Autonomous Cyber-Defense

AUTHREX-AGENT-CYBER

Autonomous cyber-defense authority governance.

After DARPA's AI Cyber Challenge (DEF CON 33, 2025), autonomous cyber-reasoning systems can patch critical-infrastructure software at machine speed. AUTHREX-AGENT-CYBER governs the missing question: may an autonomous system patch a live water-treatment or power-grid controller, at what authority tier, and with what rollback? The cyber-reasoning system is treated as a black box; AUTHREX governs the action only, with no offensive function.

Reference architecture aligned with the Five Eyes "Careful Adoption of Agentic AI Services" (1 May 2026), DARPA AIxCC, and NDAA §1513. Folds in AUTHREX-ZTAGENT and AUTHREX-MCPGOV as cited variants.

4
Traced Scenarios
4
Authority Tiers
0
Offensive Functions
LIVE
On Agent Page
◇ AUTHORIZE THE DEFENDER
AUTONOMOUS CRS"finding X, proposed patch Y" SATA · patch provenance verifiedADARA · finding/action consistencyHMAA · tier set by target criticality WHICH TARGET? TEST TARGETT3 autonomousEXECUTE LIVE OTT1 + rollbackHANDOFF POISONEDinconsistentABORT SAME PROPOSED ACTIONdifferent target → different authority FIVE EYES AGENTIC-AI GUIDANCE · DARPA AIxCCGOVERNANCE ONLY · NO OFFENSIVE FUNCTION
5-EYES · AIxCC
Agentic-AI guidance
NDAA §1513
BLADE-AGENT-HSM
◇ THE LIGHT-SPEED DELAY DECISION
GROUNDcontrol CRAFTin LEO ↔ minutes of signal delay ANOMALYstale conjunction data ONBOARD GOVERNANCE PIPELINE SATA · is the sensor data fresh and trusted?HMAA · what tier is this maneuver?FLAME · is there time to wait for ground?CARA · safe-hold if uplink never arrives T3 · safehigher tier ACT ONBOARDwithin authorityenvelope, logged GOVERNED HOLDwait for ground uplinkCARA safe-state NASA SBIR 2026 BAA · EXPAND.3.S26BSPACE POLICY DIRECTIVE 5
NASA SBIR
EXPAND.3.S26B
SPD-5
BLADE-SPACE
Federal-Need Application · Civil Space

AUTHREX-SPACECYBER

Onboard authority governance for orbital autonomy.

The same pipeline, instantiated onboard a spacecraft where ground control is light-seconds away. Signal delay removes the human from the loop, so authority must be governed onboard. AUTHREX-SPACECYBER decides what the vehicle may do for itself, at what tier, and when it must wait for a ground uplink before acting. A stale-data anomaly does not trigger an autonomous burn; it triggers a governed hold.

Reference architecture aligned with NASA SBIR 2026 BAA subtopic EXPAND.3.S26B (autonomous onboard health management for small spacecraft) and Space Policy Directive 5.

7
Pipeline Stages
4
Authority Tiers
30krad
Rad-Hard Class
LEO
Domain
Federal-Need Application · Test & Evaluation

AUTHREX-SANDBOX

Test-and-evaluation authority governance.

The same pipeline, applied not to production but to the sandbox itself. AUTHREX-SANDBOX governs what an AI under evaluation is permitted to do inside the test environment, so evaluation is bounded, audited, and reversible. Every action is capped at test-tier, no irreversible step runs without review, and the environment resets between runs.

It is the controlled setting an AI must pass through before any AUTHREX-ASSURE decision can clear it for production. Reference architecture aligned with NDAA §1534 (DoD AI sandbox-environments task force, due 1 Apr 2026) and NDAA §1533.

7
Pipeline Stages
4
Authority Tiers
TEST
Env Scoped
2
Testbed Anchors
◇ THE SANDBOX BOUNDARY
THE SANDBOX (BOUNDED) AI UNDEREVALUATION PIPELINE GOVERNS THE TEST HMAA · every action capped at test-tierFLAME · no irreversible step without reviewCARA · environment resets between runsERAM · every action logged for audit BOUNDED · AUDITEDreversible · no production reach only after passing HANDOFF TO AUTHREX-ASSUREthe pre-production gate NDAA §1534 · AI SANDBOX TASK FORCE (1 APR 2026)NDAA §1533 · ASSESSMENT FRAMEWORK
NDAA §1534
Sandbox task force
TESTBEDS
Rover + UAV
What They Share

One verified pipeline underneath all of them.

The product and all five applications above run the same seven-stage AUTHREX authority pipeline. The core is model-checked in TLA+: 48,751 reachable states verified, 8 of 9 safety properties holding, with the one known violation logged in the issue register rather than hidden. That single verified core is why these are one idea shown six ways, not six separate bets.

SATA → ADARA → IFF → HMAA → MAIVA → FLAME → CARA  ·  ERAM signed ledger
Cited Variants & Features

Not separate applications, and labeled as such.

For honesty about scope: two of the names in the AUTHREX vocabulary are cited variants folded into AGENT-CYBER, and two are features, not standalone applications. They are listed here so the count is never inflated.

Cited Variants · folded into AGENT-CYBER
AUTHREX-ZTAGENTZero Trust for autonomous agents. The same agentic surface as AGENT-CYBER, not a separate application.
AUTHREX-MCPGOVModel Context Protocol server governance. A citation layer within AGENT-CYBER.
Features · not applications
AUTHREX-PQCPost-quantum-ready signing. A property of how BLADE-AGENT-HSM signs, a hardware feature.
AUTHREX-AISBOMAI software bill of materials. A signed provenance record the ERAM ledger emits, a ledger feature.
Honest Scope

What this set is, and is not.

These five applications are reference architectures at TRL 2-4, specified, simulated, and anchored to DOI-registered hardware designs. They are not deployed, not operationally validated, and not adopted or endorsed by any agency. The mapping of each federal need to the AUTHREX pipeline is one researcher's analytical work, released openly so an independent reviewer can apply the same reasoning and report disagreement. Citations to federal guidance, law, and solicitations refer to publicly available primary sources. This is independent research offered for technical evaluation.