Alan Rozenshtein, research director at Lawfare, spoke with Miles Brundage, founding executive director of the AI Verification and Evaluation Research Institute (AVERI) and former senior advisor for AGI readiness at OpenAI, about the state of AI safety and accountability and AVERI's vision for independent third-party auditing of frontier AI companies.
The conversation covered the weaknesses of current AI regulations, including California's SB 53 and New York's RAISE Act; why Brundage left OpenAI to build an independent nonprofit; AVERI's case for shifting the unit of analysis from individual AI models to the organizations that build them; the "Volkswagen problem" of deception-proofing safety evaluations; a framework of AI Assurance Levels ranging from baseline transparency to treaty-grade verification; the limitations of safety benchmarks and the BenchRisk project's findings; market-based mechanisms for driving audit adoption, including insurance, procurement, and investor pressure; and how AVERI navigates the tension between proximity to industry and independence from it.
Mentioned in this episode:
- Frontier AI Auditing: Toward Rigorous Third-Party Assessment of Safety and Security Practices at Leading AI Companies, Averi 2026
- Risk Management for Mitigating Benchmark Failure Modes: BenchRisk, NeurIPS 2025
- Why I'm Leaving OpenAI and What I'm Doing Next, Miles Brundage, Substack, October 2024
.jpeg?sfvrsn=f8ab7a0d_7)
.jpg?sfvrsn=258e2451_5)
