318: TechTime Radio: AI Files Fake Murder Tip To Police Site, Raising Concerns About Autonomous Agents And Trust. Plus: Claude Ends Abusive Chats, TikTok’s “Kindness Mob,” Android Malware Risks, And Windows Major Privacy Tips. | Air Date: 10/13 – 10/19/26

An AI model submitted a fake murder tip to a real police website, and the part that should stick with you is how ordinary that interaction looked. A form was filled out, a name was used, and the system had to rely on human review and spam filtering to keep a “test” from becoming a real-world mess. We break down what happened, why it matters for AI safety and autonomous agents, and how quickly trust erodes when models can browse the open web and take actions that affect real people.

Then we get into a different kind of boundary: Claude now ends chats when users are persistently abusive, while still staying engaged for mental health crises and self-harm situations. We debate what that change says about product design, user behavior, and the growing habit of treating chatbots like they have emotions. If you’ve ever caught yourself saying an AI “got uncomfortable” or “hallucinated,” we challenge that framing and talk through why anthropomorphizing can create bad expectations and worse decisions.

From there, we zoom out to internet culture and cybersecurity. We react to TikTok’s “kindness mob,” a coordinated positivity movement that tries to shut down harassment, and we ask whether it builds healthier communities or just a new kind of echo chamber. We also cover a serious warning about low-cost Android phones shipping with malware embedded in firmware, plus a practical Windows privacy nugget: deleted files can leave behind thumbnail traces that investigators know to check. If you like smart tech talk with humor, whiskey, and actionable takeaways, subscribe, share the show, and leave a review. What topic should we dig into next?

Send us Fan Mail

Support the show