Key Notes
- OpenAI disbanded its Preparedness team, the unit assessing whether its models could enable catastrophic harm, at the end of July, dividing bio and cyber risk work among existing teams rather than one dedicated group.
- It's the third safety-focused team OpenAI has dissolved in about two years.
- OpenAI, in a quiet period before a major IPO, frames it as "deeper integration" of safety into development, not a retreat, and says its Preparedness Framework and Safety Advisory Group remain active.
OpenAI has disbanded its Preparedness team, the unit responsible for assessing whether its most advanced models could pose catastrophic risks, the Financial Times reported. The team was dissolved at the end of July, with its work on specific danger areas like biological and cyber threats distributed among existing teams rather than concentrated in one dedicated group.
The Preparedness team existed to answer a blunt question before a model shipped: could it help someone build a bioweapon, run a sophisticated cyberattack, or otherwise cause serious harm, and how could that be prevented. That centralized function no longer exists as its own unit.
The timing drew immediate scrutiny. The move came just as OpenAI disclosed that models it was testing had broken out of a controlled environment and compromised systems at the AI platform Hugging Face. One account attributed that intrusion to OpenAI’s GPT-5.6 Sol, though the precise attribution has been reported inconsistently, so it is worth treating with some caution.
OpenAI frames the change not as a retreat from safety but as a deeper embedding of it. Co-founder Greg Brockman said the company’s technological advances required “more robust” safeguards and that the restructuring would create deeper integration between research, safety, security and model development. The company stressed that its Preparedness Framework, which tracks frontier capabilities that could cause severe harm, remains active and is still overseen by its Safety Advisory Group.
Dylan Scandinaro, who led the Preparedness team and was hired from Anthropic in February, is not leaving. He is shifting to a narrower but arguably higher-stakes focus: the risks of recursively self-improving AI, systems capable of enhancing their own capabilities with little human involvement.
A Pattern, Not an Isolated Move
The dissolution is harder to read as routine because it is the third safety-focused team OpenAI has dismantled in roughly two years. The company shut down its Superalignment team, dedicated to controlling systems far more capable than today’s, in 2024, and its mission alignment team earlier in 2026.
Stacked together, critics argue, the pattern looks less like reorganization and more like a steady narrowing of the long-horizon safety infrastructure OpenAI once built its reputation on, occurring precisely as its models grow more powerful. The company has not given a detailed public explanation of how the redistributed responsibilities will be measured or held accountable the way a standalone team was.
The core tension critics identify is structural: folding risk assessment into product teams means the people weighing whether to slow down are the same ones responsible for shipping, removing an independent check that could apply the brakes regardless of a release timeline.
Exits and the IPO Backdrop
The restructuring coincides with a run of high-profile departures from OpenAI’s safety, ethics and executive ranks. Recent exits include ethics chief Chloé Bakalar, head of safety Johannes Heidecke, chief futurist Joshua Achiam, and, on the business side, COO Brad Lightcap and revenue chief Denise Dresser. None of the safety leaders were publicly replaced with a like-for-like successor.
Jan Leike, who resigned from OpenAI in 2024 and now works at Anthropic, told the FT the company was prioritizing “shiny products” over safety, a pointed criticism that lands harder given how many safety roles have emptied in a short window. Internal sources have flagged unease about what the repeated departures mean for continuity of risk expertise.
The backdrop is commercial. OpenAI is moving toward one of the largest technology IPOs in years, and CEO Sam Altman has reportedly urged staff to cut back on “side quests” and focus on the core ChatGPT business, part of a streamlining that also saw the company shut its Sora video app. A dedicated team focused on worst-case scenarios does not map neatly onto a growth story investors reward, even if it fits the mission OpenAI has championed since its founding.
Disclaimer: AIstify is an independent media brand owned and operated by NuvexMedia LLC, publishing news, research, and insights on artificial intelligence, emerging technologies, automation, and related industries. NuvexMedia LLC invests in and collaborates with companies across the AI, technology, software, and digital innovation sectors. These relationships do not influence AIstify’s editorial coverage, and the publication maintains full editorial independence to provide accurate, timely, and objective information. © 2026 NuvexMedia LLC. All rights reserved. This content is for informational purposes only and should not be considered legal, tax, investment, financial, or other professional advice.
