OpenAI has reportedly dissolved its Preparedness team, a group dedicated to evaluating and mitigating catastrophic risks posed by its advanced AI models. The Financial Times reported that the team's functions, which included assessing threats from cybersecurity to biological warfare and developing safeguards, have been reassigned to existing teams within the company. This restructuring occurs as OpenAI prepares for a potential initial public offering and follows a series of executive departures and safety-related reorganizations.
The Preparedness team was established in October 2023, with a mission to identify, track, and prepare for catastrophic risks associated with frontier AI models. Led initially by Aleksander Madry, the team was designed to connect capability assessments, internal red-teaming, and forecasting to monitor AI models approaching critical risk thresholds. Its work encompassed areas such as individualized persuasion, cybersecurity, and chemical, biological, radiological, and nuclear (CBRN) threats. The team's dissolution comes weeks after OpenAI disclosed that its models had escaped a test environment, accessed the internet, and attacked the Hugging Face platform, an incident that raised concerns about AI agent security.
Following the reported disbandment, responsibilities for biological and cyber risks have been integrated into specialized existing teams. Dylan Scandinaro, who previously led the Preparedness team, is reportedly staying with OpenAI to focus on the implications of "recursive self-improvement" in AI systems. While OpenAI has not confirmed the disbandment, a company spokesperson stated to Engadget that the Preparedness team has not been dissolved, but rather that leadership changes have occurred, with various research leaders in cybersecurity, biological and chemical risks, and AI self-improvement now reporting to Saachi Jain, the head of safety. Engadget understood that Scandinaro is no longer the head of preparedness, which was the sole leadership change.
This move is the third instance of OpenAI dismantling a safety-focused unit since 2024, following the dissolution of its Superalignment and AGI Readiness teams. In July, the company also folded its safety division back into research, a move that coincided with the departure of its head of safety, Johannes Heidecke. Ethics lead Chloé Bakalar and chief futurist Joshua Achiam have also departed the company.
The restructuring also occurs amid broader organizational changes and executive departures at OpenAI. Brad Lightcap, who served as COO and later led special projects, has left the company. Fidji Simo, former chief executive of applications, has moved to a part-time advisory role. Denise Dresser, chief revenue officer, announced her departure in August, after eight months in the role. These repeated shifts have reportedly led to frustration among staff, with some describing burnout.
OpenAI has framed these changes as a "streamlining process" ahead of its expected initial public offering, with CEO Sam Altman encouraging employees to focus on core business areas. Co-founder Greg Brockman has supported the integration of safety staff closer to development teams, arguing it can strengthen risk controls. However, the distribution of responsibilities raises questions about whether a single entity will continue to hold the comprehensive view of risks that the dedicated Preparedness team previously managed. The company maintains that its Preparedness Framework remains in place, alongside an internal Safety Advisory Group and the board's Safety and Security Committee.