Security expert warns AI may hack the systems society runs on, not just computers

A security author has used a keynote at the DEF CON 34 conference in Las Vegas on August 7, 2026 to argue that AI hacking has three parts: people hacking AI systems, people using AI to hack other systems, and AIs hacking systems on their own, not even maliciously. He says the last two matter more than the first, and that the third is the weirdest and potentially the most destabilising.

He says he first wrote about this in 2021 and in his 2022 book, A Hacker's Mind, when he described it as mostly science fiction, but not stupid science fiction. He now says it is being seen more and more, especially in the past year. The trends he describes do not require a major breakthrough in AI, and will happen more often as AIs get better at autonomous problem-solving. He warns that AI will find and exploit vulnerabilities in social, economic and political systems, which he calls a difference in kind, not just degree.

To make the point, he broadens the meaning of hacking. He defines a hack as something a system permits, but that its designers did not anticipate and do not want. The tax code is one example: loopholes are its vulnerabilities, and tax avoidance strategies are its exploits. He says all systems of rules can be hacked, including sports, customer reward programs, financial systems and politics. This article covers only the opening of the speech.