Vinod Khosla Warns AI Has ‘Great Uses’ and ‘Bad Uses:’ ‘AI Is Not Verifiable’

Venture capitalist Vinod Khosla argued that AI presents a unique challenge because its use can be difficult to verify, making international treaties harder to enforce.

Khosla Warns AI Is Hard to Verify

On Sunday, Khosla said on X that many technologies have both beneficial and harmful applications, citing nuclear technology and biotechnology as examples.

"AI is no different in having great uses for the bottom three billion people on the planet and bad uses," Khosla wrote.

He argued that AI presents a unique challenge because its use can be difficult to verify.

"Main difference no treaties are possible because unlike other technologies AI is not verifiable when used," Khosla said.

Khosla called on David Sacks to keep AI risks at the forefront of policy discussions.

"Up to people like @DavidSacks to keep this danger front and center," he wrote.

Khosla also cited COVID-19 as an example of the consequences of technological risks, saying it caused "7 million direct deaths and over 30 million including indirect deaths."

Many great technologies have good and bad uses like nuclear and biotechnology. AI is no different in having great uses for the bottom three billion people on the planet and bad uses. Main difference no treaties are possible because unlike other technologies AI is not verifiable…

— Vinod Khosla (@vkhosla) September 13, 2026

AI Risks and Market Concerns

Last week, the resignation of Anthropic researcher Jacob Coxon and warnings from Evan Hubinger reignited debate over AI’s potential to threaten humanity.

Hubinger estimated a greater-than-10% chance AI could kill all humans within a decade, while Geoffrey Hinton, the “Godfather of AI,” said the estimate was reasonable.

AI researcher Gary Marcus disagreed, saying there were serious AI risks but no realistic scenario he knew of that could "kill all humans."

Oxford researcher Toby Ord has estimated roughly a 10% chance of existential catastrophe from unaligned AI this century, while Roman Yampolskiy, author of “Artificial Superintelligence,” said the risk could exceed 90% if artificial superintelligence were developed.

Separately, Universa Investments founder Mark Spitznagel warned that the AI boom could eventually trigger a historic market crash after another major, euphoric rally.

He agreed with Michael Burry’s concerns over AI companies’ heavy spending but said Burry could be wrong on the timing.

OpenAI chief scientist Jakub Pachocki also urged caution, warning that increasingly capable AI agents could threaten critical infrastructure, act independently of human instructions and become harder to monitor.

He called for slower AI development and coordinated safety standards.

Disclaimer: This content was partially produced with the help of AI tools and was reviewed and published by Benzinga editors.

Photo courtesy: Shutterstock