Exclusive: Former OpenAI policy chief creates nonprofit institute, calls for independent safety audits of frontier AI models

Miles Brundage has launched the nonprofit AI Verification and Evaluation Research Institute (AVERI) to push for external auditing standards for frontier AI models and published a coauthored framework proposing “AI Assurance Levels” 1–4. AVERI has raised $7.5 million toward a $13 million target to fund 14 staff for two years and lists donors including Halcyon Futures, Fathom, Coefficient Giving and the AI Underwriting Company; it intends to research standards and policy rather than perform audits. The initiative signals a potential shift whereby insurers, large corporate customers, investors and future regulators could require independent audits—an outcome that could raise compliance costs for leading AI labs and become material in IPO, litigation and underwriting risk assessments.

Analysis

Market structure: Independent AI auditing creates clear winners—large consulting/audit firms and insurers able to provide “assurance” (Accenture ACN, Marsh & McLennan MMC, AIG) and cybersecurity/pen-test vendors (CrowdStrike CRWD, Palo Alto PANW). Frontier AI labs and small-cap pure‑play AI vendors (e.g., C3.ai AI) face higher compliance cost and time-to-market friction; incumbents (GOOGL, MSFT) gain relative pricing power because they can internalize audit overhead. A short supply of qualified auditors implies wage inflation and capacity constraints—expect 20–50% premium on specialized audit labor the first 12–24 months.

Risk assessment: Tail risks include mandatory external audits or EU enforcement that pause deployments, leading to 20–40% revenue compression for companies that monetize rapid model updates (high-impact, low-probability within 12–24 months). Short-term (days–weeks) reputational shocks for any firm losing an audit; medium-term (3–12 months) underwriting changes by insurers that force contractual audit clauses; long-term (1–3 years) standardization that raises barrier to entry. Hidden dependency: audit hiring competes with R&D/talent pools, slowing product roadmaps and possibly increasing outsourced cloud/compute demand (benefit NVDA).

Trade implications: Allocate to providers of audit/assurance and cybersecurity while de‑risking speculative AI plays. Expect consulting/audit revenue upside of +2–6% for ACN/MMC over 12–24 months and margin expansion for niche security vendors if audits become procurement prerequisites. Monitor EU AI Act compliance signals (30–90 days) and IPO filings from OpenAI/Anthropic (12–24 months) as liquidity/catalyst events.

Contrarian angles: The market underestimates recurring, high‑margin revenue from mandatory audits—historical parallel: Sarbanes‑Oxley which structurally lifted audit fees for a decade. Overdone fears: large incumbents (GOOGL/MSFT, NVDA) are unlikely to suffer lasting damage; underappreciated outcome: emergence of public audit-specialist vendors or verticalized insurers capturing outsized returns. Unintended consequence: standardization could accelerate commoditization of models, compressing pure-play multiples while boosting service/assurance revenue streams.

Key Decisions for Investors

• Establish a 1.5–2.5% long position in ACN (Accenture) with a 12–24 month horizon to capture consulting/audit revenue growth; target total return +15–25% if EU/US standards require broad external audits, stop-loss -12%.
• Build a 1.0–1.5% long position in CRWD (CrowdStrike) or PANW (Palo Alto) for cybersecurity/pen-test demand; implement a 6–12 month bull-call spread (buy 9-month ITM call, sell 9-month OTM call) to cap cost and target 25–40% upside.
• Initiate a 1.0% short or buy 3–6 month ATM puts on C3.ai (AI) as a hedge against valuation compression from rising compliance costs; cover on signs of binding regulation or if AI drops <20% from current price.
• Add a 1.0% long position in MMC (Marsh & McLennan) or public insurance broker AON (AON) to capture underwriting changes; trim if quarterly filings show no mention of AI audit underwriting within 3 quarters.
• Monitor EU conformity guidance and any SEC statements over the next 30–90 days; if EU enforcement language mandates external evaluators, scale longs above by +50% over 4–8 weeks and take profits at +20–30% or upon IPO filings from major labs (12–24 months).