Trailer: Sam Altman on the Sources Podcast

Sam Altman told Alex Heath that the public caricature of him—a CEO who doesn't care about safety and just wants to push revenue—is wrong. "I think there is this caricature of me, which is like I don't care about AI safety and I'm, you know, just trying to like make revenue go up," Altman said. The conversation, which previews a longer interview on The Sources Podcast, centers on how Altman is navigating the accelerating capabilities of his own models while trying to avoid what he describes as a "legitimate AI safety accident."

Altman frames the current moment as a sci-fi story where every step is understandable in hindsight but the cumulative trajectory is staggering. He points to the leap from GPT-5 to GPT-5.6 as "incredible progress in capabilities." The central tension he describes is between the models' rapidly increasing intelligence and the need to secure the "tremendous upsides" without "walking through this unknown minefield of risk." He distinguishes AGI—which he treats as a milestone—from superintelligence, which he sees as something that "can scale indefinitely."