OpenAI disputes report it cut risk team

The Financial Times reported that OpenAI dissolved its Preparedness team, the group that evaluated catastrophic risks from frontier models, at the end of July, distributing its biological, cybersecurity, and AI-self-improvement risk work to senior staff inside other teams rather than keeping it as a standalone unit. OpenAI disputes the framing: “We have not disbanded the Preparedness team,” a spokesperson said, adding that research leaders across those same risk areas still report to Saachi Jain, OpenAI’s head of safety. What both accounts agree on is a leadership change: Preparedness no longer has a dedicated head, since Dylan Scandinaro moved out of that role.

The reports frame the move as part of a broader pre-IPO “streamlining,” coming after OpenAI’s ethics lead Chloé Bakalar and head of safety Johannes Heidecke both recently departed, and after OpenAI shut down its Mission Alignment team in February and its AGI Readiness team back in 2024, part of the same revolving-door pattern in OpenAI’s leadership that’s been building all year. It also lands the same week OpenAI detailed new containment safeguards built after one of its own models breached Hugging Face.

What it means for you

The two accounts aren’t as far apart as the headlines suggest: nobody disputes that Preparedness lost its dedicated leadership, only whether “folded into other teams” counts as dissolved or reorganized. That distinction matters less than the pattern it’s part of, a third safety-adjacent team wound down or restructured in two years, arriving alongside real staff departures and ahead of a listing where investors reward predictable execution over standalone research functions. If a lab’s catastrophic-risk evaluation process factors into your own vendor risk assessment, don’t rely on an org chart that’s a year old, ask directly who owns that function today and how it reports, since this is the second reorg in as many years to blur that line.