Nvidia CEO Jensen Huang Calls for Shutting AI Labs That Cannot Contain Models Under Existing Liability Rules

Huang framed AI safety as an engineering and liability question during the September 2026 interview, rejecting appeals for regulatory pauses from OpenAI, Anthropic and similar firms. He noted that labs requesting antitrust and liability exemptions while expanding compute capacity hold agency to withhold unsafe products, with existing civil and criminal statutes already applicable. Huang dismissed Geoffrey Hinton's 10 percent existential risk estimate as unscientific and unsupported by prior forecasts.

The intervention highlights divergent incentives among compute suppliers and frontier labs. Nvidia gains from sustained demand without new oversight layers, while labs seek external constraints to manage competitive dynamics and potential damages. Primary records show the labs' employee letter cited internal pressure to accelerate, a claim Huang attributed to executive decisions rather than external forces.

Huang's position aligns with documented patterns where hardware leaders prioritize throughput over governance experiments. No new treaty text or legislative record supports the labs' requested exemptions, leaving liability allocation to courts under current U.S. statutes.

Next steps hinge on whether regulators treat AI outputs under existing product-liability precedents or introduce sector-specific carve-outs. Congressional records through 2025 show no enacted relief matching the labs' framing.