Y Combinator CEO: Focus on Real AI Safety Risks, Not Sci-Fi Doomsday

Y Combinator CEO Garry Tan said in a CNBC interview on Sept. 10, 2026, that too much attention is being paid to Hollywood-style AI doomsday scenarios while real risks are already emerging. "We should focus on scientific facts, not science fiction. I've seen 'Terminator 2' and it's a great movie," he said.

Tan urged policymakers to focus on concrete issues such as cybersecurity, infrastructure security, and understanding how autonomous AI agents actually operate. "If a system is compromised and multiple agents act together to take over our infrastructure, what do we do? Do we know where an agent is running? Which data center? How do we shut it down?" He cited OpenAI's intrusion into Hugging Face as a real-world example.

His remarks came as current and former researchers from OpenAI and Anthropic have publicly warned about AI risks, heightening concerns in Silicon Valley. Asked whether those insider warnings made the threat seem more credible, Tan said he believed they were likely right. "They are raising alarms, and they are probably correct," he said, adding that the public should stop fixating on vague notions that AI is evil and must be stopped.