‘We don’t need any new laws’
Nvidia chief executive Jensen Huang told an audience at Salesforce’s Dreamforce conference in San Francisco on Tuesday that artificial intelligence does not need fresh regulation, and that the companies building it should be trusted to police themselves.
“Safety is an engineering problem, not a legal one,” Huang said, according to TechCrunch, which reported his remarks. He added: “We don’t need any new laws. We don’t need new regulations.” His prescription for a company unsure about a system was simple: “If you’re not confident in its functionality, capability, or safety, then don’t release it.”
Huang also rejected the framing that speed and safety trade off against each other, calling the choice between moving fast and moving carefully a false one, and said the work that actually matters is building the test environments in which systems are assessed.
The other side of the argument was in the same building
The timing was not accidental. Dario Amodei, the Anthropic chief executive, also appeared at Dreamforce, four days after publishing an essay calling on frontier labs to slow their pace together to avoid catastrophic risk. That essay has dominated the sector since Saturday: Sam Altman, Demis Hassabis and Elon Musk have all responded to it, Microsoft’s Satya Nadella has endorsed “deliberate pacing”, and OpenAI confirmed on Monday that it has been in safety talks with Anthropic and Google for weeks.
Huang’s position is the sharpest public rejection of that argument from a company at the centre of the build-out. Nvidia sells the chips that every one of those labs trains on, and a slower frontier is, among other things, a smaller order book.
What he is and is not proposing
It is worth being precise about the claim. Huang is not arguing that AI systems are safe by default. He is arguing that the discipline that makes them safe is the same one that makes any other hardware-and-software product safe — testing, evaluation, staged release — and that a company which cannot vouch for a system should hold it back rather than being told to by statute.
The counter-argument, which the essays and the hearings of the past week have all circled, is that the incentive to hold a system back is weakest exactly when a competitor is about to ship. Huang’s answer is that the market punishes a bad release. Whether it punishes it fast enough is the question nobody at Dreamforce settled.
Where this lands in Washington
The remarks arrived on the same day that Senator Bernie Sanders and Steve Bannon shared a stage at a “Pro-Human Assembly” a few blocks from the White House, and a day before Senator John Kennedy is due to introduce a bill requiring AI developers to build shutdown mechanisms into their systems.
President Trump has spent the week dismissing the case for restraint, calling the push to slow AI a conspiracy and a hoax. On the question of new law, Huang and the administration are on the same side; on the question of whether the risk is real, they are not obviously in the same conversation.
What to watch
Nvidia reports again in November, and the gap between what its chief executive says about pacing and what its guidance assumes about demand will be worth reading side by side. The nearer test is the Senate: Kennedy’s bill, Sanders’ promised superintelligence ban and the Thune–Klobuchar work on blocking unsafe models all put the “no new laws” position on the record against something concrete.
