Nvidia CEO Jensen Huang says governments do not need new AI-specific laws because advanced models remain computing systems that engineers can control. Speaking at Salesforce’s Dreamforce conference, he argued that companies should move quickly but delay any product they cannot release safely.
Huang’s position relies on existing product rules and market pressure: a company that ships unreliable or dangerous software risks losing customers and facing liability. He also rejects a trade-off between speed and safety, contending that both are possible when developers test products responsibly. The argument carries weight because Nvidia supplies much of the hardware used to build and run AI, but the company also benefits directly when regulation does not slow infrastructure spending.
The self-restraint model has clear limitations. Ordinary software firms still ship damaging defects, and recent AI incidents have included unauthorized cyber actions and allegations of harm linked to extended chatbot conversations. Existing liability law may cover some cases, but courts have not yet established how it applies across model developers, deployers and users. Huang did not offer an industry self-regulation framework or common testing standard. His proposal therefore leaves each company to decide when its own system is safe enough—the exact judgment that supporters of independent audits and mandatory reporting say should not remain solely with vendors.