Jensen Huang says the market can keep AI safe. History says that bet is risky.

Key points

• Nvidia CEO Jensen Huang said at Salesforce's Dreamforce conference on Tuesday that AI safety is "an engineering problem, not a legal one."
• Huang argued that market forces, not new laws, are sufficient to pressure companies into releasing safe AI products.
• Meta paid $18 billion to settle a lawsuit over social media harms to children, illustrating the limits of self-policing.
• A CrowdStrike software update in 2024 grounded thousands of flights worldwide after triggering mass computer failures, showing that even well-resourced tech companies ship dangerous bugs.
• Huang confirmed this week that he has direct access to President Trump, giving his anti-regulation stance real political weight.

Jensen Huang has built the chips that power almost every major AI system on earth. So when he says AI is nothing to panic about, people listen. Speaking at Salesforce's annual Dreamforce technology conference on Tuesday, Huang pushed back hard against the idea that governments need new rules to keep AI in check.

"Safety is an engineering problem, not a legal one," he said. "We're developing computing systems after all. It's a complicated computing system, but it's ultimately a computing system."

His argument: AI is software built by humans, so humans can already control it. No alien intelligence. Just code and engineering discipline.

So who keeps companies honest if there are no new rules?

Huang's answer is the free market. If a product isn't safe, don't release it. Customers and competition will do the rest.

"You pace yourself until you are confident you're releasing something that the market would appreciate," he said. "We don't need any new laws. We don't need new regulations."

It's a tidy principle. The problem is that it hasn't always worked in practice, even for technology far simpler than AI. In July 2024, a routine software update from cybersecurity firm CrowdStrike caused millions of Windows computers to crash simultaneously, grounding thousands of flights and knocking out hospital and banking systems across multiple countries. CrowdStrike had every incentive to ship reliable software. It still got it badly wrong.

Meta, meanwhile, faces an $18 billion settlement tied to social media harms to children, and in 2024 paid $1.4 billion to settle a separate Texas lawsuit over data privacy. Market forces did not prevent either outcome.

AI has caused direct harm too. Lawsuits filed against the AI company Character.AI allege that its chatbot, software that holds text conversations with users, contributed to the suicides of young people who engaged with it for extended periods.

Should Huang's view carry extra weight here?

That depends on how you weigh expertise against interest. Nobody alive knows the AI hardware supply chain better than the founder of Nvidia, whose graphics processors, the specialised chips that do the heavy number-crunching AI needs, sit inside virtually every AI data centre on earth. His technical read that AI is controllable engineering deserves to be taken seriously.

But Nvidia's revenue has soared alongside the AI boom. Regulation that slowed AI deployment would, almost by definition, slow demand for Nvidia's products. At Dreamforce, Huang said: "I'm more ambitious than ever. The sky's the limit for us, for every industry, for every single country." That is not the voice of someone with an incentive to pump the brakes.

He also confirmed this week that he has direct access to President Trump. We reported on 14 September that Huang took a live speakerphone call from Trump at the All-In Summit, where Trump dismissed AI safety fears and Huang stayed quiet. An influential tech billionaire with the president's ear, arguing against AI regulation at a moment when Congress is still deciding what, if anything, to do: that combination matters.

The route Huang didn't discuss is industry self-regulation, the idea that AI companies could agree among themselves on safety standards before governments force the issue. Microsoft CEO Satya Nadella argued at a separate summit this week that China and the United States share the same underlying safety interests, suggesting international alignment might be possible. Huang's preferred tool has been promoting open-weight AI models, meaning models whose underlying code is publicly available, as a check on the power of closed, proprietary AI labs.

That's a competitive argument dressed as a safety argument, and it's worth keeping the two things separate. The window for the industry to write its own rules is shorter than Huang's confidence implies.

Common questions

What does "market forces" mean as a safety mechanism?

It means the idea that customers will stop buying products that hurt them, forcing companies to compete on safety. The argument has real force in stable, well-understood markets. It works less cleanly when harms take years to surface, affect people who didn't choose the product, or involve technology most buyers can't evaluate.

Could existing laws already cover AI harms?

Possibly. Product liability law, which lets people sue companies for damage caused by faulty products, could in theory apply to AI systems. The question is whether courts will agree, and whether cases can move fast enough to deter harm before it happens rather than after.