In order to trust machines with important jobs, we need a high level of confidence that they share our values and goals. Recent work shows that this “alignment” can be brittle, superficial, even unstable. In one study, a few training adjustments led a popular chatbot to recommend murder. On this episode, contributing writer Stephen Ornes tells host Samir Patel about <a href="https://www.quantamagazine.org/the-ai-was-fed-sloppy-code-it-turned-into-something-evil-20250813/">what this research reveals</a>.Audio coda from The National Archives and Records Administration.

Audio Edition: A New, Chemical View of Ecosystems

Exploring the distant universe, the insides of cells, the abstractions of math, the complexity of information itself, and much more, The Quanta Podcast is a tour of the frontier between the known and the unknown. In each episode, Quanta Magazine Editor-in-Chief Samir Patel speaks with the minds behind the award-winning publication to navigate through some of the most important and mind-expanding questions in science and math. Quanta specifically covers fundamental research — driven by curiosity, discovery and the overwhelming desire to know why and how. Join us every Tuesday for a stimulating conversation about the biggest ideas and the tiniest details.(If you've been a fan of the Quanta Science Podcast, it will continue here. You'll see those episodes marked as audio edition episodes every two weeks.)