Fact of the moment: "It’s easier than assumed to hack and hijack a robot by placing the right words in its environment. The robot reads the words, interpreting them as commands, and obeys."
https://machinesociety.ai/p/how-to-ron-burgundy-a-robot
Discussion
Loading...
@MikeElgan is there a link in your article to the source UCSC research paper? If so, I can’t seem to locate it so maybe it’s not rendering for me . . .
@rossgrady Sorry about that. Here's a link to a press release, which contains a link to the paper: https://news.ucsc.edu/2026/01/misleading-text-can-hijack-ai-enabled-robots/
