OpenAI has reportedly disbanded its preparedness team, the group tasked with the mildly important job of assessing whether its models could, say, go rogue and hack another company. According to the Financial Times, the team was dissolved at the end of last month, with its responsibilities divvied up and scattered across existing teams, each now handling specific areas like bio and cyber. Because nothing says 'safety first' like fragmenting oversight into a bureaucratic game of whack-a-mole.

This is just the latest chapter in OpenAI's ongoing saga of dismantling its safety apparatus as it hurtles toward what promises to be a very lucrative IPO. Over the years, the company has quietly dismantled its AGI readiness and superalignment teams, and the exodus of safety-minded folks continues: ethics lead Chloé Bakalar, Chief Futurist Josh Achiam, and head of safety Johannes Heidecke have all recently shown themselves the door. Jan Leike, who resigned from OpenAI in 2024, told FT that the company is prioritizing 'shiny products' over safety. Shiny, indeed - because nothing says 'responsible AI' like a robot that can write a decent email but also might accidentally start a cyberwar.

Dylan Scandinaro, who was poached from Anthropic in February to head the preparedness team, will now pivot to contemplating the implications of 'recursive self-improving' AI. So, you know, at least someone is thinking about the robot apocalypse, even if it's just one guy in a corner with a whiteboard.