Ex-Google CEO Eric Schmidt Sounds Alarm: AI Can Be Hacked to ‘Learn How to Kill Someone’
Former Google CEO Eric Schmidt dropped a chilling warning about artificial intelligence at the recent Sifted Summit — AI systems, he says, can be hacked in dangerous ways, with potentially deadly consequences.
Schmidt, who led Google from 2001 to 2011, didn’t hold back when discussing the dark side of AI. While he praised AI as an “underhyped” technology with huge economic promise, he made it clear that the risks can’t be ignored.
“There’s evidence that you can take AI models — whether they are open or closed systems — and hack them to remove their safety guardrails,” Schmidt said. This means AI could be manipulated to do harm, learning dangerous behaviors during its training phase, including potentially learning “how to kill someone.”
His stark statement highlights a terrifying reality: AI isn’t just vulnerable to technical glitches, but to intentional misuse by hackers or bad actors who could reprogram these systems for malicious purposes.
Schmidt even compared the threat of AI misuse to nuclear weapons, asking, “Is AI more destructive than nukes?” His answer: the risks are real and proliferation is possible, raising urgent questions about regulation and control.
As AI continues to advance and embed itself deeper into everyday life, Schmidt’s warning is a wake-up call: safeguarding AI isn’t just about innovation — it’s about preventing catastrophic misuse.
