Jensen Huang Wants ‘No Incidents’ in AI Cybersecurity, Tells Doomers to Quit

Nvidia chief executive Jensen Huang has set an uncompromising bar for the artificial intelligence industry’s security record: “no incidents.”

Speaking about the risks that accompany the rapid rollout of AI systems across businesses and governments, Huang framed cybersecurity as a problem the industry must solve rather than merely manage — a standard far stricter than the tolerance for occasional breaches that has long characterized enterprise technology. In the same remarks, he took aim at the industry’s most pessimistic voices, arguing that people convinced AI will end badly for humanity “ought to stop working on” the technology.

The comments land at a moment when the two themes Huang touched on — security and existential risk — have become the dominant frames for public debate about AI.

A higher bar than the industry is used to

Corporate cybersecurity has traditionally been treated as a numbers game. Organizations assume breaches are inevitable, invest in detection and response, and measure success by how quickly damage is contained. Huang’s “no incidents” framing rejects that premise for AI.

There is a logic to the stricter standard. AI systems are increasingly embedded in the infrastructure that runs power grids, financial markets, hospitals and defense systems. They are also being given more autonomy: agentic systems that take actions, write and execute code, move money and interact with other software without a human reviewing each step. A compromised model or a manipulated agent is not simply a data-loss problem — it is a system that can be turned into an instrument against its owner.

That combination of scale and autonomy is why AI security has become one of the fastest-growing areas of enterprise spending. It is also an area where Nvidia has commercial interests. The company’s dominance in AI chips has made it the foundational supplier for nearly every major model developer, and the security of the hardware-and-software stack it sells is central to its customers’ confidence.

Pushback against the doomers

Huang’s second message was blunter. The so-called “doomer” wing of the AI world — researchers and commentators who believe advanced systems pose a serious risk of catastrophic or civilization-ending outcomes — has gained prominence in recent years, influencing regulatory debates in Washington, Brussels and beyond. Some of its most visible members work inside the very labs building frontier models.

Huang’s argument is essentially one of consistency: if you genuinely believe the technology you are building will destroy the world, the rational response is to stop building it. The implicit charge is that doomerism, when voiced by people who continue to push the frontier forward, functions more as positioning than as conviction.

Critics of that view counter that safety-minded researchers stay inside the labs precisely because they want to steer the technology away from the worst outcomes — and that walking away would simply cede the field to those with fewer reservations.

Two sides of the same argument

Taken together, Huang’s comments sketch a coherent position. He is not dismissing AI risk; he is relocating it. The dangers he treats as urgent are concrete and engineering-shaped — intrusions, compromised systems, manipulated agents — rather than speculative scenarios about machine superintelligence.

It is a view that fits the optimistic, build-fast posture Huang has maintained throughout Nvidia’s ascent to the center of the global technology economy. Whether “no incidents” is an achievable goal or an aspirational slogan, it sets a benchmark the industry will now be measured against — including by the skeptics Huang would rather see leave the field. Read More