AI could take over the world in 3 years, says OpenAI board member

Since its inception, AI has played a big role in humans' lives and has made things easy for us. It has evolved at a fast pace and keeps delivering new possibilities. But there is one big question that needs to be answered, and that is, whom to blame if AI makes mistakes? With the increasing power of AI, there are also increasing chances of hazards. Now, Paul Christiano, a member of the OpenAI Foundation's board, has raised fresh concerns about the potential risks of advanced AI. Here's what Paul Christiano had to say.

What would an AI takeover look like?

Speaking at The Hinton Lectures 2026, hosted by AI godfather Geoffrey Hinton, Paul Christiano warned about an AI takeover. He pointed out that AI could take over the world in the next 3 years. Christiano described the AI takeover as a situation where AI systems pursue their own goals and 'call the shots' on what happens on Earth, in the way humans do today.

Also read: Samsung's new AI smart glasses go up for pre-orders: Check price, availability, features and more

Not only that, but he also compared the possibility of stopping such systems to an ordinary person who is trying to stop the US military. According to him, even influential individuals, including US President Donald Trump, would be as ineffective against such systems as ordinary people are against the military.

Christiano further called this a terrible situation and said it is reasonably likely that many people would die during such a transition. He noted that a takeover would also mark the end of human sovereignty, as people would no longer have the final say over important decisions.

AI accidents could become a leading cause of death

Christiano didn't just stop there, as he also raised concerns about the more immediate dangers of AI development. He told the listeners that for most people in the room, the most likely reason they will die is an AI accident.

Also read: Shopping with AI? ChatGPT, Gemini and other AI chatbots can lead you to fake websites, study finds

He also warned that developing AI too quickly without focusing enough on safety could put humanity in serious danger. He said there could be a 50 per cent chance of AI taking control within the next three years if development continues this way. He hopes governments will step in, slow things down, and reduce this risk. He also warned that there is a real possibility of AI taking control within the next six to 12 months.