Odd Lots · Thursday, September 17, 2026
Francis Fukuyama views the primary danger of Artificial Intelligence as 'agentic AI,' where humans delegate decision-making power to machines. He is concerned about the opacity of modern AI systems, stating that creators often don't fully understand their inner workings. Delegating authority to such agents could lead to a loss of human control and unpredictable, potentially dangerous outcomes.
“Well, no, I think that what I said was that the most dangerous aspect of AI is agentic AI. That is where human beings delegate to machines the power to make decisions.”
“And I think this Hugging Face incident that OpenAI got into where they had developed an AI agent that broke out of its playpen and then started getting out into the internet and breaking into other computers, that's what agentic AI is all about.”
“And the thing about modern AI is that the people that create them don't really understand what's going on underneath the hood. Classic computer programming, line by line, it's very deterministic what the computer is going to do. With these new AIs, they don't really understand why it's behaving the way it is.”