All news

24 September 2026

Palo Alto Networks' CEO says calls to slow down AI won't work

Nikesh Arora told CNBC that pacing AI development is unrealistic because not every company will hold back, and said the odds of AI wiping out humanity are far below 10%.

The debate over whether AI companies should deliberately slow down has a new voice. Nikesh Arora, who runs cybersecurity giant Palo Alto Networks, told CNBC's "The Tech Download" podcast that the idea of pacing AI progress is unrealistic, because some companies simply won't go along with it.

What happened

The argument started after an Anthropic researcher put a 10% chance on AI killing all humans, and another researcher quit, saying Anthropic and OpenAI were not acting responsibly. Days later Anthropic CEO Dario Amodei wrote an essay saying the industry should slow how fast it makes AI models more capable. OpenAI's Sam Altman, Elon Musk and Google DeepMind's Demis Hassabis voiced support.

Arora first reacted on X by suggesting the slowdown call was a clever tactical move, arguing to CNBC that AI firms might be seeking sympathy from regulators and a free pass on legal liability. He now says the bigger problem is practical. "I think not everybody is going to pace themselves," he said. If careful companies pause, less careful ones keep going, so he thinks the frontier labs should keep building but not release anything they are not confident is safe. He also noted it is unclear how anyone would measure the pace of AI progress.

Arora called the 10% extinction figure highly unlikely, describing the real risk as extremely small and saying the job is to build in safety guardrails.

Why it matters

Arora is not alone. Nvidia CEO Jensen Huang recently said new rules are not needed and that developers should test products until they are ready. Treasury Secretary Scott Bessent told CNBC that AI developers should take responsibility themselves rather than expect a federal liability shield, meaning legal protection from being sued over harm their products cause.

For investors, this is about who carries the risk. Rules that slow AI or shift liability onto developers could change how fast the industry spends and ships. Palo Alto sells security tools, and Arora has warned that AI is making cyberattacks faster and easier to scale, so it has a direct stake. It is worth about $318 billion by Noorult's data, and its shares returned roughly 93% over the past year.

What to watch

  • The proposed US-China AI Dialogue, under which, Bessent said, the US proposed a system for the two countries to notify each other of AI-related national security incidents.
  • Whether Washington moves toward rules on AI liability or leaves it to companies.
  • Whether any other AI labs publicly sign on to slowing down.

Arora doubts global coordination is close, saying the US first needs to agree on what it actually wants. Palo Alto's financials are on its Noorult company page.

The companies in this story

Based on reporting by CNBC: AI slowdown proposals ‘unrealistic': Palo Alto CEO. Facts, not advice.

More news