Foreign outlets citing five people familiar with the matter report OpenAI
employees warned management and the board before ChatGPT's launch that code and
research access was overly broad, creating risk of sensitive-technology leaks;
some employees said management prioritized catching up with Google and ANTHROPIC
over security. Cited internal risks include permissive code-repo permissions,
research materials widely stored on GitHub and Slack, and testing tools lacking
regular security reviews. OpenAI has bolstered security, reallocating roughly
25% of production engineers to a security infrastructure project in August.
OpenAI says it has tightened access to sensitive materials, increased internal
security reviews, and will launch a new targeted security initiative.
