Research
Papers and explainers from IAEF.
Key Institutional Considerations for Designing Effective Independent Verification Organizations
An independent verification organization (IVO) is a licensed, independent, expert-led body that checks whether an AI system meets specific safety, risk, and trustworthiness criteria. The idea has bipartisan support and is moving into law. But momentum is outpacing design.
Five main arguments.
- Evaluate AI where it is used. IVOs should be able to evaluate AI models, systems, and applications in context, for frontier companies and for the enterprise companies and AI-adopting organizations that deploy and use AI built by others.
- Give IVOs scoped, protected access. IVOs need information from developers and deployers, set out in a pre-evaluation plan and shared under security and confidentiality protections. Insufficient evidence should be a valid finding.
- Build the institutions, and phase in requirements. IVOs depend on certified evaluators, validated methods, and a designation & oversight process, all sitting outside the IVO itself. Requirements should phase in as evaluation science matures.
- Funding an evaluation should not buy control over it. Companies pay into a mandatory pooled fund. They do not choose their evaluator or influence the outcome.
- IVOs evaluate; others decide. IVO findings are evidence, not verdicts. Many of the bodies IVOs depend on do not yet exist, or have no IVO mandate yet.