OpenAI staff reportedly warned leaders for years about security risks

Staff reportedly warned OpenAI leaders and its board about security for years; some concluded competition mattered more than security. People familiar with the company say anyone with code access could see details of unrelated projects. OpenAI says its model weights and much of its most sensitive work are kept outside the shared repository. Thibault Sottiaux reportedly took charge of a security initiative and reassigned a quarter of production engineers in August 2026.