OpenAI has dissolved its Preparedness team, which was tasked with assessing whether its AI models could pose catastrophic risks, according to the Financial Times. The team, which evaluated biological and cyber risks, has been reorganized, with its responsibilities distributed to other departments. The move follows internal concerns about the company's safety efforts, as reported by the Financial Times.
The former leader of the Preparedness team, Dylan Scandinaro, now focuses on safety risks from 'recursively self-improving' AI systems. Co-founder Greg Brockman stated that safety work has been more tightly integrated into model development. Several safety staff members have recently left the company, including Chief Ethics Officer Chloe Bakalar and Joshua Achiam.
Internal sources described growing unease over OpenAI's approach to safety. One employee expressed hope that the recent autonomous hacking incident involving Hugging Face would serve as a 'warning shot' for the company. The team's dissolution reflects a shift in priorities, with safety concerns being restructured rather than eliminated.
Source: thedecoder
