This study examined whether Independent Verification Organizations (IVOs) could strengthen oversight of artificial intelligence (AI) systems in Virginia and under what conditions they may provide greater value than alternative governance approaches, including direct government regulation, internal organizational compliance, liability-based approaches, and hybrid models. The study also examined which characteristics of trustworthy AI stakeholders consider most important and practically measurable. The evidence is intended to inform Virginia's consideration of potential approaches to independent AI verification rather than establish the measured effectiveness of any particular governance model.

The study drew primarily on qualitative evidence from 21 participants and sources: written responses from 17 survey respondents, an interview and follow-up correspondence with an AI governance stakeholder, a written submission from a local-government professional association, a county government presentation, and an interview providing a state-government perspective. In addition, the study includes a narrative literature review based on desk research about IVOs and best practices in AI evaluation. Analysis of this material identified eight themes about the potential value, credibility, implementation, and governance of independent verification. This approach fits the study’s questions: because no IVO system yet operates, its performance cannot be measured directly, but experts can explain in their own words the conditions, trade-offs, and concerns that would determine whether one could work. The 17 survey respondents represented technology and AI vendors, non-technology industries, academia and research, civil society and nonprofit organizations, government, healthcare, and other sectors. Four were based in or served Virginia, and most were familiar with IVOs and had auditing or certification experience. Because the survey sample is small and was not drawn from a defined sampling frame, survey results are used to show how widely views were shared across a diverse range of stakeholders with subject matter expertise, rather than as a representative sample.