Chilson Proposes Federal Duty-to-Mitigate Plus Common Law for AI

On September 9, former FTC official Neil Chilson posted a detailed thread responding to economist Tyler Cowen's frontier AI regulation proposal, arguing for a federal, unified "duty of care" framework in place of Cowen's preferred industry self-regulation audits with liability exemptions. The core disagreement centers on whether passing audits should confer legal immunity, and on how safety rules should be enforced. Chilson's plan has sparked debate over regulatory costs versus safety outcomes.

Confirmed

• Chilson's central objection is Cowen's arrangement whereby passing an audit earns legal immunity: he argues this trades away the long-term benefits of common law for highly uncertain risk reduction—a deal labs would happily take, but one Congress could bargain much better on. Cowen's original proposal called for a FINRA-style industry self-regulatory nonprofit that would shield companies passing audits from liability.
• He argues Congress should establish a duty of care for frontier AI labs to "reasonably reduce the risk of mass casualties and catastrophic property loss," paired with disclosure and incident reporting requirements, and replace fragmented state AI safety laws with a single federal standard.
• On enforcement, he proposes a dual-track approach with the Commerce Department and the courts: Commerce could issue non-binding guidance and build expertise through enforcement cases for courts to draw on; halting a deployment would require persuading a court that a lab has violated, or is about to violate, its mitigation duty, rather than relying on routine administrative approvals.
• He sees common law's strength in courts evolving with evidence and ruling on specific harms and reasonable precautions; he agrees with Cowen that courts need expertise, but contends this can be addressed while preserving firms' incentives to prevent harm.
• He criticizes costly checklist-style compliance: the industry is still figuring out which AI safety practices actually work, and expensive "tick-box" rules would both push smaller competitors out of the market and dull large labs' incentives to explore genuinely effective risk reduction.

Why it matters

• The debate crystallizes the core divide in U.S. frontier AI regulation: industry self-regulation with exemption incentives, versus statutory duty of care with judicial accountability. Chilson's plan tries to avoid fragmented state laws and rigid administrative approvals while preserving the adaptability of the common law and corporate accountability, offering lawmakers an alternative, actionable legislative blueprint.

2026-09-09 ~ 2026-09-09 · 8 related posts

• Neil Chilson rejects legal immunity for audited AI labs in rebuttal to Tyler Cowen's proposal — neil_chilson · 2026-09-09
• [source] Neil Chilson Counters Tyler Cowen: No Legal Immunity for AI Labs That Pass Audits — neil_chilson · 2026-09-09
• Neil Chilson: costly check-the-box AI compliance means less competition, not safer models — neil_chilson · 2026-09-09
• Neil Chilson: common law courts can judge concrete AI harms as evidence develops — neil_chilson · 2026-09-09
• Neil Chilson: Congress Should Impose a Duty to Mitigate Frontier AI Catastrophic Risks — neil_chilson · 2026-09-09
• [source] Neil Chilson proposes a federal duty-of-care standard for frontier AI instead of licensing — neil_chilson · 2026-09-09
• [source] Chilson's Enforcement Blueprint: Commerce Guidance Plus Common-Law Accountability — neil_chilson · 2026-09-09
• A 'fourth way' to regulate AI: federal duty to mitigate catastrophic risk plus common-law accountability — neil_chilson · 2026-09-09