Structural conflict disclosure, uniform methodology commitments, and roadmap to independent governance.
The Agent Governance Standard (AGS) is published by Imperium Business Intelligence Ltd (“Imperium”), which also operates Agent Shield — an AI agent governance platform assessed on the AGS leaderboard alongside competing platforms. This is a structural conflict of interest. Imperium manages the conflict through uniform methodology, transparent disclosure, an open dispute process, paid verification with documented integrity firewalls, and a public roadmap to greater independence over time. This document describes how.
Imperium Business Intelligence Ltd is the publisher of the Agent Governance Standard (AGS) and the operator of the agentgoverning.com leaderboard.
Imperium also develops Agent Shield, an AI agent governance platform that is assessed on the same leaderboard alongside competing platforms.
This creates a structural conflict of interest that is common in technical standards work. The IETF develops the HTTP specification while many of its members operate HTTP server implementations. The Linux Foundation publishes Kubernetes alongside member companies that build managed Kubernetes services. The conflict is managed, not eliminated, through the practices below.
The scoring methodology is applied identically to all assessed platforms, including Agent Shield. Agent Shield receives no preferential treatment in:
Verified scores and estimated scores are produced by different methodologies — empirical adversarial testing for verified, documented-capability analysis for estimated — but both are presented as legitimate scores within the AGS framework. A platform with an estimated score has not been disadvantaged by choosing not to undergo verification.
Agent Shield’s score is updated under the same methodology and on the same review cadence as competitor scores. A regression in Agent Shield’s capability would be reflected in its score. Methodology changes are versioned and applied to all platform scores in the next scoring cycle, never selectively.
Agent Shield’s verified scores are produced by running the Agent Shield platform against the AGS adversarial test suite operated by Imperium. This means Imperium is both the developer and the tester of Agent Shield.
Competitor estimated scores are derived from public documentation analysis, not from running competitor platforms against the test suite. This creates an asymmetry: Agent Shield benefits from full test coverage, while competitors are scored from documentation only.
This asymmetry is disclosed on the leaderboard, on every platform page, and on every assessment publication with the labels “Verified” vs “Estimated”. Any page missing this disclosure is an error and will be corrected on report to disputes@agentgoverning.com.
Importantly, this asymmetry creates a ceiling on how meaningful Agent Shield’s lead is. Verified scores are reproducible against the AGS test suite; estimated scores reflect documented capability. A competitor that submits for verification can match or exceed Agent Shield’s score on dimensions where their platform meets the requirement, regardless of who developed the test suite. The methodology rewards capability, not authorship.
Vendors who wish to undergo independent verification — running their platform against the AGS adversarial test suite to produce a verified score — may submit for a verification engagement. Pricing, scope, and submission requirements are published at agentgoverning.com/verify.
To submit, vendors email disputes@agentgoverning.com with the subject line “VERIFICATION REQUEST: [platform name]”. Submissions should include:
Imperium maintains a limited number of complimentary verification vouchers — no more than two per quarter — allocated to platforms whose verified governance scores serve the broadest public interest. Voucher allocation is based on deployment scale and regulatory significance, not on commercial relationship with Imperium. Voucher allocations and the rationale for each are published in the AGS changelog at the time of allocation. Sponsored assessments use the same methodology as paid assessments — complimentary status does not influence scoring.
Verified scores supersede estimated scores and are labelled “Verified” alongside the score. Verification reports are shared with the vendor 7 days before publication, allowing for clarifying responses or methodology disputes raised under the Score Dispute Process. If a vendor does not respond within the 7-day window, Imperium publishes the verified score as drafted. Vendors who decline to publish a completed verification have that decision logged publicly (“Vendor X commenced verification on [date], declined publication on [date]”) to prevent score-shopping.
The fee paid by a vendor does not affect the score produced. A verified platform that fails to meet AGS dimensions receives a low verified score, published as such.
Any platform may dispute its score via disputes@agentgoverning.com. Disputes are processed under the published Score Dispute Process.
Imperium accepts payment for verification engagements at the published rate. Imperium does not accept payment for favourable scoring, score adjustments, methodology decisions, removal of disputes, or favourable timing of score updates. The integrity firewall between paid verification work and the resulting score is maintained by uniform methodology and the published dispute process. Sponsored verification vouchers are allocated according to the public-interest criteria above, not in exchange for any benefit to Imperium. Any attempt to influence a score through payment, sponsorship, or commercial relationship is logged on the public dispute register, retained permanently on the platform’s assessment history page, and may result in the platform’s score being marked “Influence Attempt Logged” for a minimum of 12 months.
AGS is licensed under CC-BY-4.0. The methodology, dimensions, and dispute process remain publicly available regardless of any future change in Imperium Business Intelligence Ltd’s ownership or governance structure. Any entity that forks or builds upon the published standard inherits the same transparency and dispute obligations documented herein.
Imperium commits to the following independence milestones for AGS:
| Milestone | Target version | Status |
|---|---|---|
| Methodology published openly under CC-BY-4.0 | AGS v2.1 | Complete |
| Test suite published publicly so any party can reproduce assessments | AGS v2.2 | Planned |
| First external assessor engaged for verification work | AGS v2.2 | Planned |
| Independent technical advisory board with three or more external members | AGS v3.0 | Planned |
| Independent dispute resolution panel | AGS v3.0 | Planned |
| Annual revenue disclosure from verification engagements | AGS v3.0 | Planned |
Status of each milestone is updated in the AGS changelog at every release. Milestones not met by their target version are explained in the corresponding release notes with a revised target.
The following table shows current accountabilities for AGS scoring work and the target separation of duties as the standard matures. All current roles are presently fulfilled by Imperium personnel; the right column shows the version at which Imperium expects to separate or supplement each role.
All current roles are fulfilled by Imperium founding maintainer Andrew Hoyle unless otherwise noted.
| Role | Currently | Target separation by |
|---|---|---|
| AGS Author | Imperium founding maintainer | Technical Advisory Board reviews changes (AGS v3.0) |
| Methodology Design | Imperium founding maintainer | Open to community RFC contributions (AGS v2.2) |
| Estimated Score Assessment | Imperium founding maintainer | External assessor engaged for verification work (AGS v2.2) |
| Verified Score Testing | Agent Aegis (operated by Imperium) | Test suite published; any party can reproduce (AGS v2.2) |
| Dispute Resolution | Imperium founding maintainer | Independent dispute resolution panel (AGS v3.0) |
| Voucher Allocation | Imperium founding maintainer | Public-interest criteria reviewed by Technical Advisory Board (AGS v3.0) |
AGS is published in the belief that independent, public, comparable governance assessment is overdue in the AI agent ecosystem. Imperium operates AGS with the recognition that comparative assessment is most valuable when it is methodologically transparent, openly disputable, and progressively more independent over time.
Imperium commits to: