OpenAI has quietly pulled apart one of its most important internal safeguards, disbanding the team responsible for spotting the most dangerous risks its artificial intelligence models could pose. According to the Financial Times, the OpenAI team disbanded at the end of July 2026 was the company’s preparedness group, the unit tasked with flagging whether its models could be misused for things like biological weapons development or large-scale cyberattacks, and figuring out how to stop that from happening.
Key takeaways
- OpenAI disbanded its preparedness team at the end of July 2026, according to the Financial Times.
- The team’s job was to assess serious risks from AI models and build ways to mitigate them, including bio and cyber threats.
- Risk responsibility is being split up and folded into existing teams rather than handled by one dedicated group.
- Ethics lead Chloé Bakalar, Chief Futurist Josh Achiam, and head of safety Johannes Heidecke have all recently left OpenAI.
- Former Anthropic staffer Dylan Scandinaro, who led preparedness, is shifting his focus to recursive self-improving AI.
What happened to OpenAI’s preparedness team
OpenAI’s preparedness team existed to answer one blunt question: could a model do real damage before it ships? That meant testing whether a system could help someone build a bioweapon, run a sophisticated hacking operation, or otherwise cause serious harm, and then designing safeguards to reduce that risk. Now that centralized function no longer exists as its own unit.
Instead, according to the Financial Times, oversight of specific danger areas like bio and cyber risk has been divided up and absorbed into existing teams elsewhere in the company. There’s no single group anymore whose entire job is watching for catastrophic misuse across the board. Whether that distributed approach catches the same problems just as effectively is something outsiders can’t easily verify from the outside.
Not the first safety team to go
This isn’t an isolated move. OpenAI has spent the past few years steadily dismantling the more research-heavy, long-horizon safety infrastructure it once built its reputation on. The company previously shut down both its AGI readiness team and its superalignment team, two groups that were explicitly focused on preparing for scenarios where AI systems become dramatically more capable than they are today.
Taken together, the pattern points toward a company narrowing its safety apparatus at exactly the moment its models are getting more powerful, not less. Each dissolution on its own might look routine. Stacked one after another, it starts to look like a strategic retreat from the kind of long-term risk planning OpenAI once championed publicly.
A wave of departures adds to the picture
The restructuring has coincided with a run of high-profile exits from OpenAI’s safety and ethics ranks. Ethics lead Chloé Bakalar, Chief Futurist Josh Achiam, and head of safety Johannes Heidecke have all left the company recently. None of them were replaced with a like-for-like successor holding the same title, at least not publicly.
That combination, a disbanded team plus departing leadership, has fed criticism from people who’ve watched the company’s safety commitments evolve over time. Jan Leike, who resigned from OpenAI in 2024, told the Financial Times that the company was ignoring safety in favor of creating what he called “shiny products.” It’s a pointed accusation, and one that lands harder given how many safety-focused roles have emptied out in a short window.
Why the timing matters
None of this is happening in a vacuum. OpenAI is moving toward what’s expected to be a massive initial public offering, and companies preparing for that kind of scrutiny tend to streamline operations, consolidate teams, and cut anything that doesn’t map cleanly onto a growth story investors can understand. A dedicated risk-assessment team focused on worst-case scenarios doesn’t necessarily fit that narrative, even if it fits the mission OpenAI has talked about since its founding.
This matters beyond OpenAI’s own walls. The company has long positioned itself as a leader on responsible AI development, and its internal choices about safety staffing tend to set a tone the rest of the industry watches closely. If a company at OpenAI’s scale is folding dedicated risk teams into product groups ahead of a public listing, competitors and regulators alike are likely to take note of what that signals about where priorities sit once commercial pressure increases.
Who’s steering risk work now
Dylan Scandinaro, who led the preparedness team before it was dissolved, isn’t leaving the company. Scandinaro, who was poached from Anthropic in February, is instead shifting his attention to a narrower but arguably higher-stakes question: the implications of recursive self-improving AI, systems capable of enhancing their own capabilities without much human involvement in the loop.
That reassignment suggests OpenAI still wants dedicated eyes on its most extreme long-term risks, just not organized the way it used to be. Whether folding bio and cyber risk assessment into existing product teams, while carving out a narrower research role for one former team lead, adds up to equivalent coverage is the open question critics like Leike are raising. For now, OpenAI hasn’t laid out publicly how the redistributed responsibilities will be measured or held accountable in the same way a standalone team once was.
FAQ
What was OpenAI’s preparedness team responsible for?
The preparedness team assessed whether OpenAI’s AI models posed serious risks, such as enabling bioweapons development or large-scale cyberattacks, and developed strategies to mitigate those risks before models were released.
Why did OpenAI disband its preparedness team?
OpenAI hasn’t given a detailed public explanation. The Financial Times reports the move happened at the end of July 2026 and comes alongside broader restructuring as the company heads toward a major expected IPO, following previous decisions to dissolve its AGI readiness and superalignment teams.
Who is handling AI risk assessment at OpenAI now?
Responsibility for specific risk areas like bio and cyber threats has been divided up and moved into existing teams rather than centralized in one group. Dylan Scandinaro, the former preparedness team lead, is now focused specifically on the implications of recursive self-improving AI.
Article produced with the assistance of artificial intelligence and reviewed by the editorial team.
