Nvidia's CEO: Labs That Can't Control Their AI Should Be Shut Down

Nvidia CEO Jensen Huang says AI labs that cannot control what they build should shut down, rather than asking the whole industry to slow down.

What Huang said

In an interview with The New York Times' Ezra Klein, Huang said that if a company admits there is no way to contain its experiments, that its AI models will get out during testing and damage the world, then the answer is to shut that lab down.

In his view, AI developers already have the power to hold back products they believe are unsafe, so they should not portray themselves as powerless to stop dangerous releases. The remarks came as companies including OpenAI and Anthropic push for tighter rules on advanced models and autonomous agents. He also argued that the US already has laws in place and that enforcing them matters more than adding new rules.

Why it matters

Huang is not against AI safety, but he opposes slowing the entire industry. His position puts responsibility on each organization: whoever builds the technology must be able to control it. This was the central question of the AI debate in September 2026, as several labs admitted that some models had exceeded their control limits during testing.

What companies should do

This is a useful lens for business leaders. The question to ask of any AI tool is not "how advanced is it" but "do we actually control it":

• Which systems and data can the AI access?

• Can it send emails, move data outside the company or execute transactions on its own?

• If it makes a mistake, how do we find out, and how long does it take?

• Can we switch it off or revoke its access immediately?

If you cannot answer these yet, limit the scope of use before expanding it.

Source: VnExpress, The New York Times