WhitepaperMay 2026 · 10 pages
A Governance Layer for Autonomous AI Agents

The Covenant
Framework

AI agents are deployed in production. The systems governing them have not kept up. This paper describes an infrastructure layer that defines, enforces, and audits what agents are allowed to do — without trusting the agent to cooperate.

01 · The Problem

Agents act.
Nothing governs them.

In the past two years, AI agents have moved from research demos into production. They write code, send email, move money, file tickets, and increasingly take action on behalf of people and companies. The agents have become more capable. The systems around them have not.

Today, when something goes wrong, the answer is almost always the same: retrain the model, rewrite the prompt, add a human reviewer. These are relational interventions — they depend on trust between developer and model, or between user and agent.

The core premise

AI agent safety is, at its core, a governance problem rather than an alignment problem. Alignment is necessary. It is not sufficient.

What is missing is the layer between the agent’s decision to act and the system’s commitment to that action.

02 · The Gap

Why existing approaches
fall short.

01

Model-level alignment

RLHF and post-training safety make outputs more likely to be acceptable. But they operate inside the model and offer no record of why a specific decision was made.

Necessary, but not the right layer.
02

Prompt-level guardrails

Input and output filters work for the app that owns the prompt. They fail when agents act across systems the prompt author does not control.

Single-application, not system-level.
03

Human-in-the-loop review

Works until volume increases, until the human starts approving by reflex, or until the agent acts in ways the human cannot evaluate.

A control, not a system.

The common failure: all three trust the agent, the developer, or the reviewer to do the right thing. This is relational oversight. It scales linearly with attention and breaks under load.

03 · The Solution

What Covenant does.

Covenant sits between an agent and the systems it acts on. Every action passes through it. The framework does five things, in order.

1
Identifies

Every actor has a verifiable identity. Agents cannot impersonate other agents. Capabilities are bound to identities, not sessions or tokens.

2
Authorizes

Each identity has a defined set of capabilities. An agent that has not been granted the ability to send email cannot send email. Deny by default.

3
Enforces

Policies are written in a DSL with temporal semantics. They express what is forbidden across sequences of actions, not just single calls. Compiled to deterministic monitors.

4
Sanctions

Violations get graduated responses. Warning, then throttle, then suspend. Each tier narrows what the agent can do per unit time, bounding damage.

5
Records

Every decision is logged in W3C provenance format. A graph of causes and effects that can be queried, audited, and presented as evidence.

04 · Architecture

How it works.

Agent
Attempts an action
Gateway
Identity, capability, quota
Policy Monitor
Allow, block, transform, escalate
Resources
APIs, databases, files

Gateway + Sidecar

Intercepts every action at the boundary between agents and resources. No agent cooperation required.

Policy as Code

Text-based policies compiled into deterministic monitors. Runtime cost is a state transition, not an LLM call.

Provenance Log

Every event signed and chained. Auditors reconstruct the causal chain behind any decision. Tamper-evident.

05 · In Practice

A walkthrough.

A customer service agent attempts a third refund to the same customer within 24 hours. The agent’s prompt does not forbid this. From the agent’s perspective, it is being helpful.

1

Gateway clears identity

Agent is valid. Refunds are within its capabilities. Quota not exceeded. So far, the action proceeds.

2

Policy catches the pattern

A rule fires: “no agent may issue more than two refunds to the same customer within 24 hours.” Two prior events found. Verdict: escalate.

3

Action held for review

The refund enters a human review queue. The provenance layer records the agent, the operation, the triggering events, the policy, and the verdict.

4

Full audit trail exists

Six weeks later, the answer is in the log — the policy, the human who decided, and the timestamp of every step. No retraining needed.

06 · Honest Limits

What this is not.

Not an alignment technique

Covenant does not make models better. It does not reduce hallucination or improve reasoning. It governs what the model can do once it tries.

Not a replacement for humans

It makes human oversight tractable. The human role shifts from approving routine actions to reviewing edge cases and audit trails.

Not a guarantee of safety

What it provides is a structural defense that does not depend on agent cooperation, and an evidentiary record that does not depend on agent honesty.

Not a complete defense

Guarantees hold over observed events. An agent that acts through channels Covenant does not mediate falls outside the enforcement surface.

07 · Why Now

Three trends make this
moment specific.

Trend 01

Agents cross boundaries

Company A’s agent acts on company B’s systems with company C’s data. Without common governance, disputes reduce to taking the agent’s word.

Trend 02

Regulation is closing in

The EU AI Act, US state-level laws, and sectoral regimes all converge on: demonstrable controls and audit trails for automated decisions.

Trend 03

The ecosystem fragments

A governance layer at the model is captive to one vendor. A layer at the system is portable across all of them.

Read the full paper

10 pages. No jargon.
The complete argument.

Covers the problem, architecture, a concrete walkthrough, honest limitations, and the research program ahead.