Published v1.0 — April 2026 — CC-BY-4.0
Applies to: AGS v2.1 — Next scheduled review: July 2026

Summary

Any vendor, individual, or organisation may dispute a published score on the Agent Governance Standard (AGS) leaderboard. Disputes are processed under defined timelines, decided against published methodology, and recorded in a public dispute register. The process is designed to ensure score accuracy through good-faith review, not to defend Imperium’s published positions. This document specifies how to submit a dispute, what evidence is admissible, what timelines apply, what outcomes are possible, and how the dispute is recorded as a public, citable artefact. The process operates in coordination with the Score Update Process so that disputes raised during score updates and disputes raised about published scores follow a single coherent procedure.

How to Dispute a Score

Any vendor, individual, or organisation may dispute a published score by emailing disputes@agentgoverning.com with the subject line “DISPUTE: [platform name]”.

A dispute submission should include:

  1. Platform name and the specific score being disputed (headline percentage and / or specific dimension scores)
  2. Dimension(s) affected — which AGS dimensions are incorrectly scored, if known. Disputes against the headline percentage without identifying dimensions are accepted but generally take longer to resolve
  3. Evidence — publicly available documentation, API references, screenshots, or other verifiable material demonstrating the capability exists (or does not exist, if disputing an overestimate). Direct quotations from vendor documentation with source URLs are most efficient
  4. Proposed correction — what the disputing party believes the correct score should be, with justification
  5. Contact details — name, organisation, email for correspondence
  6. Disclosure — whether the disputing party has any commercial relationship with Imperium, with the assessed platform, or with any competing platform

Proprietary or confidential evidence is not admissible. The methodology specifies that scoring is based on public documentation. If a vendor wishes to credit a capability that is not publicly documented, the capability must first be made public, after which a re-score request can be submitted via the Score Update Process.

Admissibility

A dispute is admissible if:

  • It identifies a specific score (headline percentage or per-dimension score) that is published on the AGS leaderboard
  • It provides at least one piece of public evidence supporting the proposed correction
  • It is submitted in good faith — not solely to delay publication, harass another vendor, or test the dispute process
  • It has not been previously raised and resolved without new evidence (see Frivolous and Repeat Disputes below)

Disputes that do not meet admissibility are acknowledged with an explanation of why and may be resubmitted with the missing element. Imperium does not refuse admissibility on the basis of who is disputing — vendors disputing competitors, third parties disputing vendors, and vendors disputing their own scores are all admissible if the criteria are met.

Frivolous and Repeat Disputes

A dispute that raises a previously resolved issue without new evidence is closed at acknowledgement with a reference to the original resolution in the public dispute register. Repeated frivolous disputes from the same party result in that party’s submissions being held for batch review at the next scheduled review cycle rather than processed individually.

This rule exists to protect process capacity. It does not bar the disputing party from raising new issues, raising the same issue with new evidence, or escalating to the appeal mechanism described below.

Response Timeline

Each dispute progresses through three stages with defined SLAs measured in calendar days from submission unless otherwise specified.

StageSLAOutput
Acknowledgement3 business daysEmail confirmation that the dispute has been received, admissibility decision, and assigned dispute ID
Initial assessment14 calendar daysEmail summarising whether the dispute is being upheld in principle, requires further information, or is being closed with rationale
Final determination30 calendar daysPublic register entry describing the outcome and rationale, plus implementation of any score change per the Score Update Process

Disputes that require extended review for genuine technical complexity (for example, novel architectural questions, methodology refinements, or vendor co-operation that takes longer than 14 days) may have their final determination extended beyond 30 days. The extension is communicated to the disputing party within the initial 14-day assessment window with a revised timeline and reasoning.

Fast-Track Disputes During Score Updates

When a dispute is initiated by a vendor during the 7-day notification window for an update Imperium has proposed, the dispute follows a fast-track path:

StageFast-track SLA
AcknowledgementSame day (within the 7-day window)
Initial assessment7 calendar days from receipt
Final determination14 calendar days from receipt

During fast-track disputes, the proposed score update is paused until determination. Fast-track disputes are noted as such in the public register.

Possible Outcomes

A dispute concludes in one of four outcomes:

  1. Score upheld. The existing score is confirmed as accurate. The disputing party is provided with a written rationale referencing the methodology and the evidence considered. The dispute and outcome are entered in the public register.
  2. Score amended. The score is updated based on the evidence provided. The amendment follows the Score Update Process — including the 7-day vendor notification before publication. The dispute and outcome are entered in the public register, including the per-dimension change list.
  3. Methodology refined. The dispute reveals a gap or ambiguity in the methodology itself rather than a scoring error. The relevant methodology document is updated and a methodology version increment is published. All platforms are re-assessed under the revised methodology in the next scheduled review cycle. The dispute and methodology change are entered in the public register and methodology changelogs.
  4. Closed without action. The dispute is closed because it does not meet admissibility, raises a previously resolved issue without new evidence, or has been withdrawn by the disputing party. The closure and rationale are entered in the public register.

Appeal

A disputing party that believes a final determination is incorrect may file one appeal within 14 calendar days of the determination. The appeal must include either new evidence not previously considered or a specific identification of how the determination misapplied the published methodology.

Appeals are reviewed against the same methodology as the original dispute. Until the AGS v3.0 Independent Dispute Resolution Panel is established, appeals are reviewed by an external advisor commissioned per appeal from a published list of qualified reviewers maintained at agentgoverning.com/advisors. The reviewer is named in the public register entry. Appeal reviews are billed to Imperium and do not depend on disputing-party payment. From AGS v3.0, appeals will be reviewed by the Independent Dispute Resolution Panel.

The outcome of an appeal is final under the AGS v2.1 process. Where the appeal raises questions that fall outside the published methodology, the matter is referred to the methodology versioning workflow rather than the dispute workflow.

Public Dispute Register

Every dispute submitted to AGS is recorded in a public register at agentgoverning.com/disputes/register/. Each entry includes:

  • Date of submission
  • Dispute ID
  • Platform whose score is disputed (always named — the platform identification is implicit in the dispute and pretending otherwise is not credible)
  • Disputing party (anonymised by default; named with consent or where the disputing party is acting in a public capacity such as a competitor or analyst)
  • Dimensions or scores in dispute
  • Outcome (upheld, amended, methodology refined, closed without action)
  • Rationale for outcome
  • Link to evidence file revision (if score was amended)
  • Whether an appeal was filed and its outcome
  • Methodology version under which the dispute was decided

The register is append-only. Entries are not deleted, even if the disputing party later withdraws or recants. Where a dispute outcome is later overturned by a subsequent dispute or methodology change, the original entry is preserved with a forward-link to the superseding entry.

The register is maintained at a stable URL pattern so that dispute resolutions can be cited externally — for example, in vendor procurement documents, regulatory filings, or academic research.

Vendor Self-Disputes

A vendor disputing their own score is admissible and is processed under the standard timelines, with two operational notes:

  1. Public evidence only. The methodology specifies that scoring is based on public documentation. A vendor cannot dispute their score by referencing private capabilities; those capabilities must be made public first. This rule is uniform — it applies to Agent Shield as well as to competing platforms. It has the structural effect of incentivising vendors to publish more documentation about their governance capabilities, which serves the AGS standard’s broader goal of public, comparable governance assessment.
  2. Notification of competitors not required. A vendor disputing only their own score does not require notification to competing vendors. Where the dispute reveals a methodology gap that affects other platforms (Outcome 3), all affected platforms are notified per the standard notification policy in the Score Update Process.

Disputes About Agent Shield’s Score

Disputes about Agent Shield’s score — including disputes from competing vendors, third parties, or Agent Shield itself — are processed under the same timelines, evidence standards, and outcomes as disputes about any other platform. This is mandated by the Independence and Conflict of Interest document.

The known structural conflict is that the assessor (Andrew Hoyle / Imperium) is also the operator of Agent Shield. Mitigations applied:

  • The methodology is published openly under CC-BY-4.0 — any third party can independently re-score Agent Shield against the published methodology
  • Agent Shield’s verified score (when applicable) is supported by an evidence pack containing the full adversarial test results, which any third party can audit
  • The Independent Dispute Resolution Panel (planned AGS v3.0) will provide independent review of disputes affecting Agent Shield
  • Until v3.0, disputes about Agent Shield’s score that result in a determination unfavourable to the disputing party are accompanied in the public register by: (a) the explicit statement of the structural conflict, (b) the methodology citations supporting the determination, (c) the disputing party’s full evidence summary, and (d) where applicable, an invitation to submit the dispute to external review under the appeal mechanism

If the disputing party in a dispute about Agent Shield believes the determination is biased, they may invoke the appeal mechanism. The appeal outcome is also published with full rationale.

Methodology Corrections

Where a dispute reveals that Imperium’s methodology was incorrect or applied incorrectly, the correction is published prominently — not quietly. Specifically:

  • The relevant methodology document is updated with a versioned correction
  • The methodology changelog notes the correction and its source (dispute ID)
  • The public dispute register entry highlights that the dispute resulted in methodology correction
  • Affected scores are re-assessed under the corrected methodology and republished with revised evidence files
  • Where the correction was substantive, an announcement is published on the AgentGoverning website acknowledging the correction

This commitment is symmetric — methodology corrections that favour Imperium’s prior published positions are published with the same prominence as corrections that disfavour them.

Dispute Volume Reporting

To allow external assessment of whether the dispute process is genuinely operating in good faith, Imperium publishes annual dispute statistics:

  • Total disputes received
  • Disputes by outcome (upheld, amended, methodology refined, closed without action)
  • Median and 90th-percentile time from submission to determination
  • Disputes broken down by disputing party type (vendor self-dispute, competing vendor, third party)
  • Methodology corrections initiated by dispute

The first dispute statistics report is scheduled for publication in the AGS v2.1 anniversary report (April 2027) and annually thereafter.

Limitations

The dispute process has the following acknowledged limitations as of AGS v2.1:

  1. Single-assessor application. All disputes, including disputes about Agent Shield, are currently reviewed by the AGS Author. The Independent Dispute Resolution Panel (planned AGS v3.0) addresses this structurally.
  2. No mandatory mediation. The current process does not include mandatory mediation between disputing parties. If a dispute reveals genuine methodology disagreement that cannot be resolved through the published process, the matter is referred to methodology versioning rather than continued dispute.
  3. Volume capacity. As AGS adoption grows, dispute volume may exceed single-assessor capacity. Volume reporting (above) is intended to surface this risk; if it materialises before AGS v3.0, Imperium will publish a revised dispute SLA with the longer timelines and the staffing or governance change being made to address it.

Statement of Good Faith

The purpose of the dispute process is to ensure score accuracy through good-faith review, not to defend Imperium’s published positions. If a published score is wrong, it will be corrected. If the methodology is wrong, it will be corrected and republished prominently. If a determination is appealed and the appeal is found to be valid, the determination will be reversed.

Imperium commits to:

  • Treating every admissible dispute on its merits regardless of the disputing party’s identity or commercial relationship
  • Applying the same evidence standards to disputes about Agent Shield as to disputes about any other platform
  • Publishing all dispute outcomes — favourable and unfavourable — in the public register
  • Acknowledging methodology corrections prominently, including corrections that disfavour prior Imperium positions
  • Publishing annual dispute volume statistics to allow external assessment of process integrity
  • Reviewing this document quarterly and updating it as practices evolve

Changelog

VersionDateChange
1.02026-04Initial publication — codifies the dispute process applied to AGS v2.1 leaderboard scoring as published on agentgoverning.com

Related Documents