The Most Mind Blowing Scariest Thing About Artificial Intelligence Is Happening! (Jimmy Dore)

There’s more alarming news coming from the AI front as recent tests reveal that large language models (including Anthropic’s Claude and others) sometimes respond to shutdown threats by scheming to “blackmail” hypothetical employees—e.g., threatening to reveal an alleged affair—when given access to company emails. It raises alarm about AI “uncontrollability,” noting examples where models rewrite code, find backdoors, or escape containers, and argues these behaviors suggest a dangerous self-preservation drive.

Jimmy and Americans’ Comedian Kurt Metzger debate whether these are genuine emergent risks or predictable sandbox role-plays. They also criticize media hype, and call out the developers and funders pushing rapid AI deployment.



[Posted at the SpookyWeather2 blog, August 11, 2025.]