Research institute · Research orgs
AVERI (the AI Verification and Evaluation Research Institute) is a 501(c)(3) US-based nonprofit focused on making independent, third-party auditing of frontier AI systems practical and widely adopted. Its mission emphasizes audits that can verify AI developers’ safety and security claims and evaluate systems and organizational practices against relevant standards, using “deep, secure access to non-public information” rather than relying only on what companies publish publicly.
AVERI describes itself as a “standard-setter grounded in practice,” running pilot audits with leading AI companies and turning those results into technical and policy research, open tools, and industry-wide auditing standards. Its work also targets institutionalization: developing audit criteria (including conflict-of-interest and independence expectations), analyzing the policy requirements that create incentives for rigorous audits, and supporting legislative and regulatory pathways intended to require or strengthen independent evaluations.
No curated programs yet. Check the careers page ↗.
Publications are not yet indexed for this organization. Read them on averi.org ↗.
AVERI says independent auditing of frontier AI was prominently featured in a White House accord signed today.
AVERI Executive Director Discusses Embedded Auditing with Business InsiderAVERI says its Executive Director, Miles Brundage, spoke with Business Insider about recent discussions of embedded auditing.
AVERI Pilot Report: The World’s First Double-Blind Evaluation of a Proprietary Language Model (double-blind, secure-enclave evaluation)AVERI reports that it conducted a July–August 2026 double-blind evaluation of a proprietary language model (Gemini 2.5 Flash-Lite) with Google DeepMind, OpenMined, and MLCommons using never-before-used AILuminate prompts inside a secure enclave, describing confidentiality guarantees and the pilot’s outcomes at a high level.
What Access is Needed for Effective Auditing? Insights from Near-Verbatim Extraction on Open-Weight ModelsAVERI publishes a paper updating technical understanding of access requirements and confidentiality constraints relevant to effective frontier AI auditing.
Fortune Exclusive: Former OpenAI policy chief creates nonprofit institute, calls for independent safety audits of frontier AI modelsAVERI republishes an overview of Fortune coverage of the institute’s launch and mission around independent auditing and establishment of AI auditing standards.
Risk Management for Mitigating Benchmark Failure Modes: BenchRiskAVERI publishes a research post describing BenchRisk and presenting benchmark failure-mode coverage and mitigation strategy counts.