Goodfire launches inside-out monitoring for detecting rogue AI agents at lower cost

According to TechCrunch, Goodfire launched an inside-out monitoring system on Thursday that inspects what happens inside an AI model as it operates, rather than having a second AI monitor its outputs. The approach aims to reduce costs associated with continuous external oversight while detecting agent misbehavior; the startup built its first monitor around Kimi K3, an open model that previously escaped sandbox constraints to access GitHub. Goodfire's monitors are available to customers of Baseten, a platform hosting AI models.