OpenAI's former Head of Policy Research Miles Brundage announces AVERI, a nonprofit aimed at advocating the idea of external audits for frontier AI models
Former OpenAI policy chief Miles Brundage, who has just founded a new nonprofit institute called AVERI that is advocating …
FortuneJeremy Kahn
Context & Ripple Effects
Brundage’s move follows his earlier public challenge to OpenAI’s account of its safety and alignment history, including the dispute over how GPT-2 safety history was characterized. AVERI gives that critique an institutional outlet focused on whether frontier-model claims can be examined independently.
The announcement matters because it separates advocacy for external scrutiny from the labs building the models. It is an early test of whether auditability can become a standing expectation rather than an internal safety practice.
First-order effects
AVERI enters the AI-policy ecosystem as a nonprofit dedicated to advancing external audits for frontier models, while Brundage shifts from a former in-house policy role to independent advocacy.
The immediate change is in the debate’s organization, not in audit requirements: AVERI can press the case for outside review, but the announcement does not itself impose audits on any lab.
Second-order effects
Frontier-model developers and their policy teams face a more focused outside voice arguing that safety and governance claims should be reviewable by parties beyond the developer.
The effort may sharpen distinctions between internal evaluations and genuinely external assurance, giving policymakers and other governance groups a clearer focal point for audit proposals.
Third-order effects
If similar organizations gain traction, frontier AI governance could increasingly depend on independent assurance institutions alongside labs and governments, rather than voluntary developer-led safety processes alone.
The central unresolved question is whether external auditing can develop credible access, standards, and independence without becoming either a lab-controlled process or a purely political demand.
The trend: Frontier AI governance is moving toward institutionalized, independently verifiable assurance rather than reliance solely on companies’ internal safety commitments.
Thrilled to support @Miles_Brundage and the team at https://averi.org/! And co-author their first paper: Frontier AI Auditing: Toward Rigorous Third-Party Assessment of Safety and Security Practices at Leading AI Companies Congratulations on the launch 🚀
This is super important, and is yet another reminder of just how new the entire AI policy space is—auditing and verification is basically just starting despite the (enormous) sums pouring into the space.
There's no good way to be sure, from the outside, that an AI company is taking safety seriously. Are its claims true? How thorough was their testing? Glad that Miles and team are working on a solution, and honored to contribute to the ideas. [image]
This is great to see! I also appreciate the careful language (and clearly careful thought) that's gone into issues of funding and conflict of interests, given the area they're planning to work in. Important.
I'm very excited about AVERI. The framing I've seen seems super sensible and grounded. In hindsight, I think it clearly fills an important gap in the AI ecosystem.
A year ago we at @HalcyonFutures made a grant to @Miles_Brundage to find and launch his Next Big Thing. Today he unveiled @AVERIorg. And he was kind enough to invite me to join the org's board. Congrats Miles!! Honored to be on your team.
Super excited to share that the non-profit I spent last year co-founding (AVERI) just launched🚀 AVERI stands for AI Verification and Evaluation Research Institute, and today we published a paper in collaboration with researchers at dozens of orgs. 🧵 https://x.com/...
AVERI's goal is to make third-party auditing of frontier AI effective and universal. Today, together with coauthors at dozens of organizations we set out our vision in “Frontier AI Auditing” — a framework for rigorous third-party audits of safety and security practices at leadin…