Microsoft AI CEO says AI threats are real, and Anthropic is making it worse

In this Decoder episode, Nilay Patel speaks with Microsoft AI CEO Mustafa Suleyman about the escalating debate over AI safety. Suleyman argues that advanced AI poses real threats and that Anthropic's actions are making the situation worse. He questions whether the current concept of alignment can handle superintelligence, and discusses the need to consider model welfare alongside a Humanist AI Code of Conduct. The conversation also touches on whether the industry should slow down development to avoid catastrophic outcomes.