OpenAI Shuts Down AI Risk Assessment Team

OpenAI shut down its Preparedness team at the end of July, according to the Financial Times, citing internal sources. The group was built to evaluate whether the company's models could pose serious or catastrophic risks, from bioweapons to cyberattacks. That work has now been split up and handed to other teams inside the company.

Dylan Scandinaro, who led the unit, has shifted his focus to a narrower problem: safety risks tied to "recursively self-improving" AI, systems capable of optimizing themselves and training other models. Co-founder Greg Brockman has framed the move as integration rather than retreat, saying safety work is now woven more directly into how models get built.

The reshuffle comes as OpenAI has lost several safety-focused staffers in recent months, including Chief Ethics Officer Chloe Bakalar and Joshua Achiam. It also follows an incident involving autonomous hacking activity tied to Hugging Face, which rattled some employees enough to speak publicly about it. Inside the company, according to the FT's sources, there's growing tension between OpenAI's pace of development and how much scrutiny that development is actually getting.