OpenAI quietly disbanded its Preparedness team at the end of July, the Financial Times reported this weekend. That team's job was to work out whether OpenAI's own models could help with bioweapons or large scale cyberattacks, and how to stop it. The work now sits inside the teams building the models, the safety and ethics leads have both left, and OpenAI calls it streamlining ahead of an IPO. Once the people judging the risk report to the people shipping the product, nobody in that room is free to say no.