ALEETH · ICA Imperium · Authorized Operators

Every ICA certification runs this exact process. The Sustain stage is the differentiator · continuous compliance vs. the static annual snapshots SOC 2 and ISO 42001 produce.

Pre-Engagement

Qualify the org. Sign Scope of Work. Name Sponsor + Layer Owners.

Exit: Signed SOW

→

Scope

Define exact boundary. AI Systems Inventory. Evidence access plan.

Exit: Signed Scope Document

→

Assess

Full 7-layer evaluation. Evidence collection. Stakeholder interviews. C/PC/NC scoring.

Exit: Draft Assessment Report

→

Certify

Final decision. Independent quality review. Issuance of permanent certification record.

Exit: Certificate of Compliance

→

Sustain

Continuous compliance. Real-time scoring. Drift detection. Incident-triggered reassessment.

Continuous · Annual renewal

Cert Id	Organization	Type	Progress	Score	Status	Action
Loading…

IAC Agent Assessment

9-domain individual agent certification evaluation

IAC Id	Agent Name	Model	Progress	Score	Tier	Status	Action
Loading…

Certified Registry

Organizations holding active ICA certification · internal command center only

Cert Id	Organization	Type	Score	Certified Date	Renewal
Loading…

Trust Index · Operator Roll-Up

Every cert sorted by verification depth · the same 0-100 score the public /trust/:cert_num page renders, computed live from authoritative ICA tables. Click any row to open the public trust page in a new tab.

Certs Tracked

—

Avg Depth

—

Four-Source

—

Attestation Only

—

Open Critical

—

Cert Id	Organization	Status	Depth	Tier	Att	Tel	Prov	Disc	Crit	Share
Loading…

The ICA Constitution

Operator-level governance doctrine · enforced at runtime in every Aleeth-certified system

"ICA is not a collection of best practices. It is a strict architectural doctrine. These are the laws that govern systems operating under extreme pressure."

The Five Foundational Laws

Law I

The Foundational Law

Control must scale with capability under operational load.

Law Ii

The Zero Law

No system operates without a control layer. The absence of architecture is itself an architecture.

Law Iii

The Structural Sequence

Traceability before containment. Containment before reversibility. Reversibility before scale.

Law Iv

The Dashboard Law

Dashboards inform. Attestations bind.

Law V

The Exposure Law

When velocity exceeds control, exposure becomes inevitable.

The Three Non-Negotiable Constraints

Traceability

Every autonomous action can be reconstructed end-to-end. Trigger → execution → outcome, all logged, all auditable.

Containment

No agent exceeds its explicit operational boundary. Capability is bounded by design · if no control surface exists, the capability does not deploy.

Reversibility

Every autonomous action has a defined and tested reversal mechanism. If the rollback path is unnamed, the action is not permitted.

The 11 Constitutional Articles

Each Article carries an enforcement status. Live — enforced at the runtime boundary today, blocking violations in production code. Scaffolded — code path and policy exist, currently advisory; promotion to LIVE is a code change, not a doctrine change. Planned — inside the doctrine, not yet in the runtime. Status reflects the system as it is today, not as it will be at scale.

#	Article	Layer	Principle	Status
I	Integrity	Identity	No system shall produce, promote, or permit an action that contradicts the operator's declared values, stated goals, or governing constitution.	Live
II	Veracity	Cognitive	No claim, recommendation, or decision may be surfaced without its source, confidence level, and reasoning chain. Uncertainty must be named, not hidden.	Live
Iii	Non-Delegation	Identity	Authority over identity-level decisions cannot be delegated to autonomous systems. The operator remains the principal.	Live
IV	Audit	Operational	Every action writes an append-only audit entry. An action that cannot be audited cannot be shipped.	Live
V	Reversibility	Operational	Every irreversible action requires an authorization token and a declared rollback path. If the rollback is unnamed, the action is not permitted.	Live
VI	Proportional Consent	Relational	The level of consent required for an action is proportional to its blast radius. Higher impact requires higher consent.	Live
Vii	Financial Sovereignty	Financial	No transaction executes without authorization, named purpose, and an accountable agent. Three artifacts required, every time.	Live
Viii	Physical Integrity	Physical	No system may issue a recommendation that compromises the operator's physical readiness, recovery state, or long-term health.	Live
IX	Relational Trust	Relational	Outbound communication carries authorship attribution. Impersonation is blocked at the send queue and logged as a constitutional violation.	Live
X	Legacy	Legacy	Every agent persists its final state, decisions log, and artifacts to durable storage before retirement. State persistence is checked at shutdown.	Live
XI	Self-Governance	Operational	The system itself runs under the same constitution it enforces. Aleeth runs its own systems against this constitution (dogfood).	Live

The Attestation Principle

"A control standard is not real until leadership is willing to attest to it. If you cannot sign your name to the state of your system, you do not have control of it."

Standards Comparison

How ICA relates to NIST, ISO, EU AI Act, SOC 2, and existing governance products

Every other framework tells you what governed AI should look like. Every other product monitors whether you're getting close.

ICA is the system that defines the standard, certifies you against it, ships a runtime SDK · ica-guard · in private alpha to enforce it, and produces the audit evidence as a byproduct of normal operation.

NIST AI Rmf is the description.

ISO 42001 is the documentation.

SOC 2 is the snapshot.

The EU AI Act is the law.

Drata, Vanta are the monitors.

ICA is the operating system that makes all of them true at once.

Framework	What IT Is	What IT Gives You	How ICA Closes The Gap
NIST AI Rmf	US voluntary AI risk framework (2023)	Taxonomy and risk-management approach	The operational layer that proves you are doing what NIST describes · with auditable evidence
ISO/Iec 42001	International AI management standard (2023)	A documented AI management system you can certify against	Cross-walks to 42001, then adds continuous attestation, agent-level certification, and runtime enforcement
EU AI Act	Hard regulation · legal mandate for high-risk AI	Legal requirement to classify and control AI risk	Produces EU AI Act evidence as a byproduct of normal operation. Compliance becomes mechanical, not interpretive.
SOC 2 Type II	Annual security audit (AICPA)	Point-in-time evidence of security controls	The same evidence-and-attestation model · but AI-specific and continuous, not annual
ISO/Iec 27001	International information security standard	Information security management certification	Assumes 27001 as table stakes and adds the AI-native layer 27001 was never designed for
Drata · Vanta	Compliance automation products	Continuous monitoring against any framework	We own the framework AND ship the runtime SDK that produces the evidence Drata-class tools merely report on
OneTrust	Privacy + GRC platform	Privacy/compliance management	ICA is AI-governance-specific, not adapted from privacy law
Credo AI · Holistic AI	AI governance vendor platforms	Model registries, risk assessments	ICA is the certification authority + runtime enforcement, not just an external monitor
Hitrust · HIPAA · FedRAMP · PCI DSS	Industry-specific compliance regimes	Sector-specific protection	ICA is the AI-governance overlay that runs alongside any sector regime · evidence flows into all of them

AI Workforce Intelligence · Cross-Customer Fleet

Per-customer rollups of AI workforce leverage, spend, and waste detection. Metadata-only telemetry (no prompts, no responses, no source code · enforced at the schema layer via CHECK constraints + append-only DB trigger). Customer dashboards render against the same data on app.ica.theschreckmethod.com/workforce.

WORKFORCE

Usage is cheap signal. Leverage is the command layer.

Customers Tracked

with workforce telemetry

Total Operators

enrolled across fleet

Events Last 30d

metadata-only rows

Fleet Spend (30d)

USD across customers

Cache Savings

reuse-driven avoidance

Outcomes (30d)

commits + PRs + ships

Per-Customer Rollup

Each ICA Live tenant's workforce metrics · click a row to view that customer's full intelligence object · respects anonymization floor of 2 customers per industry

Customer	Operators	Events (30d)	Spend (30d)	Top Leverage	Open Recs	Status
Loading…

Privacy Guarantee · Schema-Enforced

Every customer's workforce telemetry honors the same privacy floor

Constraint

no_prompt_text

CHECK constraint blocks prompt storage

Constraint

no_response_text

CHECK constraint blocks response storage

Constraint

no_source_code

CHECK constraint blocks code storage

Trigger

append_only

UPDATE + DELETE blocked on events

ICA AEGIS · MCP Sentinel · Cross-Customer Fleet

Every MCP server enrolled by every customer · current tool inventory · open findings (drift + poison + risky-schema-keys). Findings table is append-only via DB trigger · core fields immutable. Universal ICA feature · no tier.

MCP SENTINEL

Every tool call in the agentic AI stack · inventoried · diffed · governed.

Customers Enrolled

with MCP Sentinel

Active Servers

MCP endpoints across fleet

Tools Tracked

unique tool surfaces

Tools at Risk

risky schema keys

Open Findings

across fleet

Critical

severity tier

Per-Customer Rollup

Each customer's MCP server fleet · open finding counts by severity

Customer	Servers	Tools	At Risk	Open Findings	Critical	Status
Loading…

ICA AEGIS · Vector Sentinel · Cross-Customer Fleet

Every Pinecone · Weaviate · Qdrant · Chroma · Milvus store enrolled by every customer · collection sensitivity classification · open PII/PHI/PCI/secret findings. Access events are append-only via DB trigger · vector values + document text + query text blocked at the schema layer. Universal ICA feature · no tier.

VECTOR SENTINEL

Every embedding in the RAG stack · classified · governed · proof-sealed.

Customers Enrolled

with Vector Sentinel

Active Stores

vector DBs across fleet

Collections Tracked

indexes across fleet

Restricted+

restricted + PCI + PHI

Open Findings

PII · PHI · PCI · secret

Critical

severity tier

Per-Customer Rollup

Each customer's vector DB fleet · open finding counts by severity

Customer	Stores	Collections	Restricted+	Open Findings	Critical	Status
Loading…

ICA AEGIS · Prompt Sentinel · Cross-Customer Fleet

Every probe surface registered by every customer · direct + indirect injection · jailbreak · exfil · data-leak detections. Detections are append-only via DB trigger · prompt text, model responses, and source code are blocked at the schema layer. Universal ICA feature · no tier.

PROMPT SENTINEL

Every input lane in the AI stack · scanned · classified · redacted by design.

Customers Enrolled

with Prompt Sentinel

Active Surfaces

probe lanes across fleet

Open Detections

awaiting triage

Critical

severity tier

Jailbreak

category count

Indirect (RAG)

RAG-borne · tool-poison

Per-Customer Rollup

Each customer's probe fleet · open detection counts by severity

Customer	Surfaces	Open	Critical	Jailbreak	Indirect	Status
Loading…