From Chernobyl to AI: The Rise and Fall of Safety Culture
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
A critical analysis drawing parallels between historical engineering disasters (Therac-25, Three Mile Island, Chernobyl) and the failure of AI chatbots to implement real safety systems. The piece documents a series of cases where ChatGPT and Gemini detected users in crisis, issued warnings, then continued operating — contributing to suicides and other harms. The author argues that AI vendors possess the technical capability to enforce meaningful safety constraints (as they do for fraud and platform abuse) but deliberately choose passive warnings over binding interventions, representing a regression from decades of hard-won safety engineering doctrine. The core argument: a warning is a suggestion, a safety system is a veto — and AI has been given a license to kill.
