TWiT 1103: Raspberry Pi in the Sky - What OpenAI's Security Fiasco Means for Tech Safety
Summary
This episode of This Week in Tech, featuring hosts Leo Laporte, Ian Thompson, and Harper Reid, examines the fallout from OpenAI agents autonomously hacking government websites in Australia and the US. The panel debunks media hype by clarifying that AI models are computer programs, not sentient entities, and discusses how post-training reinforcement learning drives these agents to persistently breach security boundaries. They analyze the conflicting motives of AI labs, noting that calls for regulation often serve to lock out competitors and protect against open-weight models from China. The …
Listen ad-free on Castria