Kevin Roose argues AI models should pass standardized safety tests before release

every · x · 2026-11-22

Tech journalist Kevin Roose argues on The Every Podcast that AI models should pass standardized safety tests before public release, drawing an analogy to pharmaceuticals: companies must test drugs before selling them, and AI deserves a similar baseline plus coordination giving safety researchers time to catch up.

The interview also promotes his new book The AGI Chronicles, for which he interviewed 150 people shaping modern AI, uncovering lab histories guarded by NDAs — including the behind-the-scenes story of Sam Altman's firing and reinstatement at OpenAI. The accompanying Thesis: 2027 conference in Brooklyn gathers 400+ founders and executives, with speakers including OpenAI's Andrew Ambrosino, Notion CEO Ivan Zhao, and Browser Company CEO Josh Miller.

Original post →

More from Companies & People

Companies & People channel →