Making frontier AI auditing effective and universal

Safety and security claims about AI systems are largely self-reported by companies. AVERI exists so that those claims can be independently verified and companies can be evaluated against common standards. 

Our flagship report, coauthored with dozens of leading experts, lays out the vision for third-party auditing of leading AI companies.

Our work makes that vision operational. We conduct research and carry out pilot audits in order to inform audit standards, policy requirements, and tooling. Our first pilot findings will be published in Q3 of 2026. 

Mature industries like aviation, finance, and nuclear power offer valuable lessons in the risks of self-reporting and the power of external oversight. We believe frontier AI demands a similar approach.