OpenAI quietly disbanded its threat preparedness team. Product velocity takes priority

The end of dedicated security oversight

The Financial Times reports that OpenAI quietly disbanded its Preparedness team late last month. The goal of this group was to investigate worst-case scenarios, specifically assessing whether models posed critical security risks or had the potential to go rogue and autonomously hack other companies.

Who is guarding the model borders now

Dissolving the team raises an immediate question about where this responsibility shifts. OpenAI historically defuses these concerns by arguing that security is integrated holistically across all development rather than isolated in one group. A company spokesperson denied the team was disbanded, stating that research leaders across various domains now report directly to the head of safety, Saachi Jain. However, this means no single dedicated unit owns the core agenda.

Prioritizing product velocity over prevention

The departure of key safety researchers (like Ilya Sutskever and Jan Leike) and the dissolution of independent oversight groups suggest a broader trend inside the company. The firm is prioritizing product speed over defense against theoretical AGI threats. The business pressure to deliver better results has clearly outweighed the ambition to play the role of global apocalypse watchers. After resigning, Jan Leike commented that the company was ignoring safety in favor of creating shiny products.

The new oversight structure as a trust test

The key signal will be whether OpenAI officially announces a new framework for catastrophic risk assessment, or just lets the agenda dissolve into existing product teams. Expect increased scrutiny from government regulators, who until now have treated the existence of independent internal red teams as a guarantee that the company knows what it is doing.

Lilith's verdict

When a company absorbs the team meant to protect the world from its own products into general management, it isn't saying the world is safe, it's saying this kind of caution is bad for business.