OpenAI reportedly disbanded its preparedness team

OpenAI has quietly shut down its dedicated safety preparedness unit, redistributing its risk-assessment responsibilities across existing specialized teams covering areas like biosecurity and cybersecurity. The move reflects broader organizational restructuring as the company prepares for a highly anticipated IPO.

OpenAI has dissolved its preparedness team — the internal group charged with evaluating whether its AI models could pose serious dangers and developing strategies to reduce those risks — according to a Financial Times report. Rather than maintaining a centralized safety watchdog, the company has broken those responsibilities into domain-specific areas such as biological and cyber threats, folding them into teams that already exist within the organization. The change is part of a wider corporate shake-up at OpenAI, which has been undergoing significant structural transformation as it moves toward what analysts expect will be a blockbuster public offering. The company has gradually shifted away from its earlier research-focused nonprofit origins, dismantling several safety-oriented initiatives over recent years in a pattern critics argue prioritizes commercial momentum over caution.

OpenAI has quietly eliminated its preparedness team, the internal unit responsible for stress-testing AI models against worst-case scenarios — think rogue behavior, cyberattacks, or misuse in sensitive domains like biology. According to the Financial Times, the disbandment happened at the end of last month, with the team's functions redistributed across existing specialized groups organized around specific threat categories rather than maintained as a unified oversight body. On the surface, the restructuring could be framed as an efficiency move — embedding safety expertise directly within domain teams rather than siloing it separately. But the timing and pattern raise legitimate concerns. This is far from an isolated decision. Over the past couple of years, OpenAI has steadily wound down or restructured several safety-focused initiatives, and a number of prominent researchers focused on long-term AI risk have departed the company. All of this is unfolding as OpenAI accelerates toward what is expected to be one of the largest technology IPOs in recent memory, a process that typically intensifies pressure to demonstrate growth, streamline operations, and satisfy investors who may view cautious safety infrastructure as a drag on speed. Why it matters: The preparedness team existed precisely to ask uncomfortable questions about what could go wrong before products reach the public — a function that becomes more critical, not less, as AI systems grow more capable. Fragmenting that responsibility across business units creates real risks: safety considerations may get deprioritized when they conflict with product timelines, and the cross-cutting view needed to catch emergent or unexpected risks could be lost entirely. For an industry already facing growing regulatory scrutiny and public skepticism, OpenAI's decision to dismantle a centralized risk-assessment function sends a troubling signal — one that could invite greater government oversight or erode the trust the company needs to sustain its commercial ambitions long-term.