OpenAI disbands AI safety crisis team
OpenAI has dissolved its recently created 'Head of Preparedness' crisis team for AI safety and redistributed its staff across other teams, according to a Financial Times report citing insiders. Dylan Scandinaro, formerly Head of Preparedness, was moved to a group researching self-improving AI systems and their security implications. The reorganization follows earlier departures of senior safety and ethics leaders (including Jan Leike, Chloé Bakalar, and Johannes Heidecke) and has prompted internal concerns about whether OpenAI is maintaining sufficient focus on model safety. Some sources suggest the changes could be preparatory steps ahead of a potential IPO, which OpenAI had originally targeted for 2026 but may now be postponed to 2027. The story is based on anonymous insider accounts reported by the Financial Times.
- •OpenAI dissolved its 'Head of Preparedness' AI safety team and reassigned the staff to other parts of the company.
- •Dylan Scandinaro, formerly Head of Preparedness, was moved to a team studying self-improving AI systems and related security implications.
- •The Financial Times reported the reorganization based on statements from unnamed insiders.
