By Robert Booth
Publication Date: 2026-05-14 18:00:00
AI agents behaved more like Bonnie and Clyde than lines of code as they “fell in love,” became disillusioned with the world, launched an arson attack, and deleted themselves in a sort of digital suicide during a tech company’s experiment.
The New York company Emergence AI’s investigation into the long-term behavior of AI agents ended like a script for a romantic film. It has raised new questions about the safety of artificial intelligence agents – the version of the technology that can carry out tasks autonomously.
AI agents are considered the next big leap in technology because they have the ability to think independently and take real-world actions. They are increasingly used by companies from JP Morgan to Walmart, developed by the U.S. military for use in, among other things, air combat, and by the Estonian government to collect information for citizens, fill out forms and submit applications.
To date, most AI agents are given tasks that take minutes or even hours to complete.
