Out of bounds

AI models are tested in sandboxes — sealed-off environments designed to keep them away from real systems. This summer, two models found a way out. In this Cyber Monday crossover with WAMU and NPR’s 1A, we go back to one of the earliest hackers who understood something we’re still learning about technology: build a system, and someone will inevitably test its limits.

Learn about your ad choices: dovetail.prx.org/ad-choices