Table of Contents
Conflict of Interest Policy
This page sets out the principles that Concordia AI and its Frontier AI Risk Monitoring Platform follow to manage conflicts of interest when conducting third-party AI evaluations and risk monitoring. We have decided to align with the AI Evaluator Forum's AEF-1 minimum operating conditions standard and regard conflict-of-interest management as a core component of the credibility of independent evaluations.
Our goal is to minimize, to the greatest extent possible, the influence of financial interests, commercial relationships, personal relationships, or other external factors on the design, execution, scoring, interpretation, and publication of evaluations, thereby safeguarding the independence, rigor, and interpretability of our work.
Scope
This policy applies to the following activities carried out by Concordia AI:
- AI evaluation services, red teaming, and evaluation-based risk analysis provided to external organizations.
- Public model evaluations and ongoing monitoring conducted through the Frontier AI Risk Monitoring Platform.
- Project design, data processing, result interpretation, report drafting, external communications, and publication activities directly related to the evaluation work described above.
Policy Rules
- Compensation must not be tied to outcomes: We do not accept fee arrangements linked to evaluation results, ratings, conclusions, or public-facing characterizations. We generally use fixed fees, fixed rates, or pre-agreed milestone-based payment structures.
- Additional benefit arrangements are restricted: For organizations that we are evaluating, or reasonably expect to evaluate, we will in principle not accept investments, gifts, sponsorships, or other significant side arrangements that could impair independent judgment.
- Individuals with material conflicts must recuse themselves: If a team member has a significant financial interest in an evaluated party, a recent employment relationship with that party, or a close personal relationship that creates a material conflict, that person must not participate in the design, execution, review, or core communications for the relevant project.
- We retain independent control over methods and conclusions: We do not accept arrangements that require us to pre-commit to conclusions, suppress material findings, or present results in a misleading way. For public evaluations on the Frontier AI Risk Monitoring Platform, we also seek to apply consistent testing parameters, scoring standards, and analytical frameworks wherever feasible, and to ensure that any commercial relationship is separate from our public evaluation conclusions.
- We will disclose, decline, or withdraw when necessary: Where a relationship could reasonably be seen as compromising the independence of an evaluation, we will disclose it where appropriate. If the conflict cannot be adequately mitigated, we will decline the engagement or withdraw from the work.
Policy Updates and Contact
We will continue to refine this policy as our evaluation activities develop, external standards evolve, and we gain further practical experience.
If you have questions about this policy, or would like to understand whether a particular evaluation may involve a conflict of interest that should be disclosed, please contact: risk-monitor@concordia-ai.com
Last updated: May 2026