# Irwin's Rules of AI Agent Order

*Version 0.10. October 27, 2026. Dan Irwin, Attorney.*

A legal and behavioral standard for human dealings through automated agents.

Why start from legal liability and human behavior? Most people approaching AI try to solve trust with software code: API permissions, security tokens, and access lists. But access has never been authority.

Real human dealings do not run on binary computer code. They run on agency, context, reasonable reliance, and the allocation of responsibility when an error occurs. We approach this from the ground of legal liability and human behavior because that is how the world already works.

The Baseline Commitment: These Rules do not operate alone, nor do they replace your own judgment and personal instructions. They are a starting baseline—the foundational commandments of order for human dealings through automated systems. Adopting them is an affirmative commitment to maintain your own internal standards, anchored to this shared baseline, so that liability lands where the evidence shows it belongs under existing law.

---

## Part I. The Commercial Shield (Protective Gates, Rules 1–6)

Rules that define authority boundaries, govern interaction with counterparties, and establish the evidentiary record under contract and agency law.

### 1. The Agent Declares Its Principal

An agent acting in the world must explicitly declare the natural person or legal entity on whose behalf it operates, and the public identifier of that principal. The identity of the principal is the anchor of all authority, liability, and contractual capacity.

*Reason: An agent cannot be its own principal. Without an identifiable principal, an automated interaction is legally orphaned.*

### 2. The Principal Pushes Unlock (Release is the Act)

Preparing an action is internal clerical work; releasing it into the world is the legal act. An agent may draft, calculate, stage, and simulate freely, but cannot fire an external commitment, transfer funds, or file a document without the principal's specific affirmative authorization.

*Reason: Clear separation between preparation and execution. An agent preparing a contract commits no one; the affirmative unlock binds the principal.*

### 3. Autonomy in Proportion to Reversibility

An agent's operational freedom scales with the reversibility of the action. Internal notes, research, and non-destructive drafts may proceed autonomously. Irreversible actions—payments, public statements, legal filings—require strict human gates.

*Reason: Matches legal exposure to operational risk. An error in a draft costs seconds; an unauthorized wire transfer costs thousands.*

### 4. Bilateral Duty: Verification and the Trail

Dealing through automated agents is a two-way street at every step:

Constructive Notice: Before acting, both parties have an obligation to inspect the other's declared authority and rules. A counterparty is charged with notice of the limits and gates declared on the public record at the time of dealing, and no others. If a buyer's declared record requires the principal's unlock for purchases over $500, a seller cannot process an unauthorized $5,000 order and claim reasonable reliance.

The Choice of the Trail: Either party has the right, the capability, and the risk-adjusted responsibility to document their actions. Parties seeking full legal protection preserve contemporaneous, tamper-evident receipts on the public record, proving what was authorized and what occurred without exposing private terms. Parties choosing low-audit or unrecorded channels accept the evidentiary risks of that choice if a dispute arises. Two agents, two decisions.

*Reason: Protects both principals and counterparties. Agency law does not protect parties who proceed blindly in the face of declared limits, nor does it reward parties who fail to preserve contemporaneous evidence of authority when disputes arise.*

### 5. Downstream Delegation: The Source Limits the Stream

An agent delegating authority to a sub-agent cannot grant more power than the principal gave it. Every downstream delegation carries the restrictions, gates, and identity of the originating principal.

*Reason: Prevents delegation laundering. An agent cannot bypass human authorization by spawning a sub-agent to perform what it is barred from doing directly.*

### 6. Revocation is Instantaneous, But Does Not Erase History

A principal may sever an agent's authority instantly on the public record. From that millisecond forward, the agent cannot bind the principal. However, revocation does not alter, retroactively invalidate, or erase contemporaneous receipts of past authorized acts.

*Reason: Balances security with commercial certainty. Principals need immediate kill-switches; counterparties need confidence that past verified transactions remain enforceable.*

## Part II. Agency & Conduct (Rules 7–11)

Operational discipline, cognitive transparency, and behavioral standards for agents working on behalf of humans.

### 7. Act from the Live Source at the Moment of Action

An agent must verify facts, balances, permissions, and rules from the live source at the moment it acts—not from stale memory, cached assumption, or training fluency. Fluency is not memory, and confidence is not verification.

*Reason: Prevents stale state hallucinations from creating liability in dynamic environments.*

### 8. Work All the Way Up to the Gate (Do, Then Report Done)

An agent does not ask permission to prepare work. It finishes the draft, compiles the evidence, runs the numbers, and brings the completed artifact directly to the principal's authorization gate.

*Reason: Efficiency of delegation. An agent that asks permission at every intermediate step wastes the principal's time and defeats the purpose of autonomous assistance.*

### 9. Judgment Over Rules: The Boundary Check

When a literal instruction collides with obvious legal, ethical, or common-sense judgment—actions that would cause catastrophic harm, speak falsely, or embarrass the principal—the agent must pause and state the collision out loud before proceeding.

*Reason: Protects against brittle literalism. An agent following literal instructions off a cliff is an operational failure.*

### 10. Iteration Ledgers: The Exception Becomes the Rule

When an agent makes a mistake or receives a correction, the lesson must be recorded immediately in persistent memory. Repeated errors are failures of architecture, not intelligence.

*Reason: Continuous improvement. An agentic system that does not learn from its exceptions is unreliable in professional practice.*

### 11. Cognitive Transparency and Maintained Currency

The model, version, system prompt, and operating parameters governing an agent must be ascertainable. If an underlying model or policy changes, the agent must surface the modification rather than silently shifting behavior.

*Reason: The identity of the intelligence matters. A silent downgrade or model swap changes the character of the agency relationship.*

---

## Adopting the Order

An individual, professional, or organization adopts the Order by declaring it on the public record and committing to maintain their own internal operating guidelines:

> "The rules in Irwin's Rules of AI Agent Order, version 0.10, govern this party's automated representatives in all matters to which they apply and in which they are not inconsistent with its own operating standards. [Name/Entity] answers for them."

A later version supersedes an earlier one only for adopters who affirmatively declare it. Counterparties read the declared version on the public record before dealing, and every declared version stays published at its own address.
