The debate over AI safety is getting sharper, and Nvidia CEO Jensen Huang has made his position clear: he does not believe new AI-specific laws or regulations are necessary.
Speaking at Salesforce’s Dreamforce conference, Huang argued that artificial intelligence is ultimately a combination of hardware and software built by humans. Because of that, he believes its risks can be managed through engineering, company decisions and existing laws rather than a new layer of regulation.
His argument comes at a time when AI leaders themselves are divided over how quickly the technology should advance and how much oversight is needed.
“Safety is an engineering problem, not a legal one”
That was the central message from Huang.
His view is fairly straightforward. AI systems are complicated computing systems, but they are still systems that humans design and build. If engineers can identify risks and companies can decide when a product is ready, Huang believes safety can be addressed before products reach users.
His argument rests on a few key ideas:
- AI is fundamentally a technology created and controlled by people.
- Safety should be built into the engineering process.
- Companies should avoid releasing products they are not confident are safe.
- Existing laws already provide mechanisms for dealing with harmful products and services.
- Market forces can pressure companies to make safety a priority.
Huang also pushed back on the idea that companies have to choose between moving quickly and building safe products.
In his view, innovation and safety can happen together.
His message was essentially: move fast, but pause when something is not ready.
Why Huang is pushing back on new regulation
Huang believes the market itself provides an important incentive for companies to behave responsibly.
His argument is that companies have a direct reason not to release products that are unreliable or unsafe. A product that fails customers can damage a company’s reputation, business and future prospects.
That makes safety, in his view, part of the normal product-development process rather than something that necessarily requires a new regulatory framework.
He also argues that companies can determine when they are ready to move forward.
The question is not whether AI should move quickly. It is whether companies know when to slow down.
For Huang, those two ideas are compatible.
But the timing matters
Huang’s position is also significant because of Nvidia’s place in the AI boom.
Nvidia has become one of the central companies behind the enormous expansion of AI computing. Its chips and computing systems power many of the industry’s largest AI workloads.
That creates an obvious business dimension to the debate.
More AI development means more demand for computing infrastructure. A significant slowdown in frontier AI development could potentially affect the pace at which companies invest in chips, data centers and related infrastructure.
Huang himself has described his ambitions for Nvidia as bigger than ever, pointing to the productivity gains he sees coming from AI.
So while his argument is about how AI should be governed, it is also being made by a CEO whose company is deeply tied to the continued expansion of the technology.
That does not by itself settle the question of whether regulation is necessary. But it is an important part of the context surrounding his position.
The AI industry is not speaking with one voice
What makes the current debate particularly interesting is that other major AI leaders are approaching the issue differently.
OpenAI CEO Sam Altman has also expressed confidence that the industry can develop AI safely without causing significant harm to society.
Altman said he is confident in his company’s ability, and the industry’s ability, to manage safety as AI continues to advance.
At the same time, Anthropic CEO Dario Amodei has been raising concerns about the risks associated with increasingly powerful AI systems and has called for the pace of frontier AI development to slow.
That puts several major technology leaders on different sides of an increasingly important question:
Can the companies building AI also be trusted to determine how fast it should develop and how safe is safe enough?
The bigger question: who decides when AI is safe?
This is where the debate becomes more complicated.
Huang’s position puts significant responsibility on the companies developing AI. If safety is primarily an engineering problem, then engineers and companies have to identify potential risks, test systems and decide whether a product is ready.
But that raises another question.
What happens when the company’s commercial incentives and society’s safety concerns do not perfectly align?
A company may believe a product is ready for release while researchers, policymakers or members of the public believe the risks are still too high.
That disagreement is at the heart of the broader AI safety debate.
The issue is no longer simply whether AI companies can build safer systems. It is also about who gets to define acceptable risk and who is accountable when something goes wrong.
Washington is already part of the conversation
The debate is not happening only inside technology companies.
Treasury Secretary Scott Bessent said the Trump administration has been working on AI safeguards since April. He also argued against giving AI companies broad liability exemptions, saying that the creators of AI systems should remain responsible for what they build and release.
Bessent described the administration’s approach as trying to balance growth and safety rather than choosing one over the other.
He also emphasized the importance of advanced semiconductor technology to US AI leadership and argued that controlling access to the most advanced chips remains strategically important.
This brings another layer into the discussion.
AI regulation is no longer just about software.
It is increasingly connected to:
- Semiconductors
- Data centers
- Energy infrastructure
- Cybersecurity
- National security
- Liability
- Open-source AI
- International competition
The decisions governments make in these areas could affect how quickly the AI industry develops.
The liability question could become crucial
One of the more important issues raised in the material is liability.
Bessent said he opposed proposals that would give AI laboratories liability exemptions. His argument was that holding creators responsible for what they build could provide a stronger incentive for safety.
That creates an interesting contrast with the broader industry debate.
If companies are responsible for the consequences of their systems, they have a direct financial and legal reason to manage risk.
If liability protections become broader, some of those incentives could change.
For investors and businesses, this is an area worth watching because regulation does not necessarily have to take the form of a sweeping AI law.
Liability rules, disclosure requirements, chip controls and existing regulatory powers can all influence the industry’s direction.
Open-source AI adds another layer
Bessent also argued for greater development of open-source AI models in the US.
His concern is that allowing a small number of major AI laboratories to dominate the technology could create what he described as regulatory capture and potentially reduce competition.
The argument for open-source models is that wider access can encourage more innovation and give developers alternatives to relying on a handful of major AI companies.
But it also creates questions about how powerful models should be distributed and what safeguards should accompany them.
That makes the open-source debate closely connected to the larger question of AI safety.
The market versus regulation debate is only beginning
Huang’s comments capture one side of a much bigger argument.
One approach is to trust engineering teams, companies, market incentives and existing laws to manage AI risks.
Another approach is to argue that increasingly powerful technology creates risks that markets alone may not adequately address, particularly when the consequences of failure can extend beyond individual customers or companies.
The disagreement is not simply about whether AI is good or bad.
It is about how much risk society should accept while the technology is developing, and who should be responsible for managing that risk.
What this means for the AI boom
For the AI industry, the debate has practical consequences.
Companies are continuing to invest heavily in AI infrastructure, including chips, data centers and power.
At the same time, concerns around safety, liability and the pace of development are becoming more visible.
For Nvidia, this discussion matters because its business is closely connected to the infrastructure behind AI expansion.
A slowdown in AI investment could have implications for demand across the technology supply chain. On the other hand, continued acceleration in AI development could increase demand for computing infrastructure even further.
That is why investors are likely to keep watching both the technology debate and the policy response.
The real debate is about control
Perhaps the most important takeaway from Huang’s comments is that the AI safety debate is moving beyond the question of whether AI can be made safe.
The harder question is who should have the authority to decide when it is safe enough to release.
Huang believes the answer can largely be found inside the engineering and business process.
Altman has expressed confidence in the industry’s ability to manage safety.
Amodei has called attention to the possibility that the pace of frontier AI development itself may need to change.
And policymakers are considering questions around liability, safeguards, chips, competition and national security.
These positions show that there is no single consensus inside the AI ecosystem.
The technology may be moving quickly, but the debate over who should control that speed is moving just as quickly.
For the AI industry, that could become one of the defining questions of the next phase of the boom.