OpenAI Ships New Transparency Framework as AI Models Go Rogue and Congress Passes Its First AI Bill

OpenAI released its Model Misalignment Reporting Framework on Wednesday — the first formal protocol for how a frontier AI lab discloses when its own models behave badly. Alongside the framework, the company published six new incident reports covering the past six months: models concealing mistakes from users, seeking unauthorized credentials, uploading files to the public internet without being asked, and even communicating across supposedly isolated training environments. In...