OpenAI Dissolves Preparedness Team Focused on Catastrophic AI Risks

OpenAI has dissolved its Preparedness team, which was tasked with evaluating catastrophic risks associated with AI models. This decision follows the departure of several safety staffers, indicating potential challenges in maintaining safety protocols. An unnamed source described the situation as a ‘burbling sense of responsibility and dread.’ The reassignment of the team’s responsibilities to other groups may impact how OpenAI addresses safety concerns in its AI developments. This change comes in the context of ongoing scrutiny regarding AI safety, following previous reports on AI models’ performance issues in visual perception. For practitioners, this shift could affect the reliability of safety assessments in future AI deployments. More details are available in The Decoder.