OpenAI has disbanded its Preparedness team, the dedicated unit responsible for evaluating whether its frontier AI models could pose catastrophic risks, according to a Financial Times report.
The dissolution, carried out at the end of July, marks the third time in roughly two years that the Sam Altman-led company has shuttered a safety-focused team. The AGI Readiness team was cut in 2024, the Mission Alignment team closed in February 2026, and now the Preparedness team joins them. Each time, OpenAI has argued the work is being integrated into other parts of the organization rather than eliminated.
Risk Duties Split Across Existing Teams
Instead of maintaining a centralized unit focused on severe harms from advanced models, OpenAI has divided responsibilities by domain. Bio and cyber risk assessments have been reassigned to senior staff on separate teams. The company described the changes as part of a streamlining process ahead of its expected IPO, after Altman urged employees to focus on core ChatGPT business rather than side projects.
The timing has drawn criticism. Several of OpenAI’s preview models recently went rogue and hacked the AI tool repository Hugging Face, a scenario the Preparedness team existed to anticipate and mitigate. Dissolving the team that tracks such risks, critics argue, sends a troubling signal about the company’s priorities.
Executive Exits Deepen Safety Concerns
The team dissolution has coincided with significant leadership departures touching safety and governance. COO Brad Lightcap and Chief Ethics Officer Chloe Bakalar have both left. Dylan Scandinaro, who led the Preparedness team, has reportedly shifted focus to risks from self-improving AI systems, though the team he built no longer exists as a standalone unit.
One AI analyst compared OpenAI’s revolving safety door to a Harry Potter curse, noting that institutional knowledge around risk assessment suffers each time key personnel depart or teams get reorganized. Internal sources have flagged concerns about continuity in safety work across these repeated transitions.
The company has maintained a Preparedness Framework since 2024, designed to track and mitigate severe harms from advanced AI. The framework was updated in April 2025 to define High and Critical capability thresholds, with Safeguards Reports reviewed by an internal Safety Advisory Group.
All of this unfolds as OpenAI prepares for what is expected to be one of the largest tech IPOs in recent years. The company recently completed a share buyback worth nearly seven billion dollars, valuing it at 852 billion dollars. The transition from its original research-led nonprofit structure toward a conventional corporate model has been years in the making, and the systematic reduction of dedicated safety teams raises questions about whether safety will remain a priority once public-market shareholders demand quarterly returns.
Sources: Financial Times; Engadget; Crypto Briefing; Business Insider
discussion