OpenAI just made an intriguing move. According to the *Financial Times*, the Preparedness team, which was responsible for assessing catastrophic risks of advanced models, was disbanded at the end of last month. Its members and responsibilities have been redistributed across departments like safety systems and research. This team originally focused on evaluating extreme risks such as cyberattacks, biochemical threats, manipulation capabilities, and AI going out of control. The contrast lies in the fact that back in May, OpenAI publicly stated that the Preparedness Framework was still the foundation for managing severe risks associated with cutting-edge AI. Disbanding the team doesn’t mean stopping safety evaluations, but it does signify that safety work has shifted from being a centralized, independent function to being further integrated into product and research systems. As OpenAI begins preparing for an IPO, the real question isn’t whether they’re still talking about safety, but who can still say “no” before the launch button is pressed. https://www.(theverge.com)/ai-artificial-intelligence/980817/openai-disbands-preparedness-team