In this excerpt from a CNN-News18 debate held on Sept 14, 2026, I explain why AI’s greatest near-term danger is not (yet) an “evil syknet machine” but powerful, out-of-control, unaligned autonomous agents making catastrophic mistakes without context, accountability or supervision.

I believe we are entering what I call an AI emergency. As someone who has worked on this for 10 years and written books on the subject, I am not a doomer, but to me this feels like a real crisis.

When industry leaders like Sam Altman, Elon Musk, and Dario Amodei publicly acknowledge serious AI risks, I view them a b it like Frankenstein’s ‘father’ getting cold feet once they see the result – although they are driven by potential trillion-dollar public offerings, no one wants to be remembered as the originator of a global purge of humanity by AI.  If a major incident triggered by their AI products occurs a few months down the road, it would doom their stock. That makes AI security an existential issue for their survival.

Also, as Nobel laureate Geoffrey Hinton noted, AI is grown, not programmed. Much like a child that grows into an adult doing whatever he/she wants, we simply do not know how to control it when it grows up. I foresee a major, disastrous AI event in the next 6 to 12 months, whether that is a financial market meltdown, an air traffic control system crash, or an energy grid failure. This will not occur because AI is inherently evil or holds malicious intent, but because it would be totally misaligned with its assigned mission. For example, a misaligned AI agent focused solely on stopping global warming might cut off oxygen without understanding the broader consequences for human life.