OpenAI has officially disbanded its Preparedness team, according to reporting from the Financial Times. The dedicated unit was tasked with proactively stress-testing frontier models against catastrophic threats, tracking automated cyberattacks, and establishing firm mitigation frameworks before deployment. Following the dissolution, OpenAI fractured these responsibilities, dispersing narrow domains like biosecurity and cybersecurity across siloed, linear product teams.
This structural pivot completes the dismantling of OpenAI's independent safety architecture ahead of an expected multi-billion-dollar initial public offering. The dissolution follows the earlier disbandment of both the Superalignment and AGI Readiness teams, accompanied by a wave of high-profile departures including safety head Johannes Heidecke, ethics lead Chloé Bakalar, and chief futurist Josh Achiam. As former alignment lead Jan Leike noted after his own resignation, safety culture has consistently taken a back seat to shipping "shiny products." Meanwhile, former Anthropic hire Dylan Scandinaro, who led Preparedness, has been reassigned to study recursive self-improving systems.
For enterprise architects, CTOs, and tech leaders deploying OpenAI APIs and autonomous agents, the takeaway is stark. OpenAI's pivot from centralized risk containment to rapid commercial release cadence effectively shifts the burden of rigorous safety auditing, defensive red-teaming, and vulnerability mitigation directly onto corporate engineering teams.