OpenAI employees had warned company leaders and board members about weak internal security for years, Reuters reported on November 3. The concerns began before ChatGPT launched in late 2022, when staff feared outsiders could access sensitive artificial intelligence research.
Five people familiar with the company’s practices said employees questioned broad access to internal code and research documents. Some workers believed OpenAI prioritized competition with Google and Anthropic over stronger security protections.
The warnings resurfaced after OpenAI disclosed in October that investigators found a recording device inside an executive conference room.
US authorities charged former OpenAI engineer Abdulrahman Alfozan in connection with the recording device. Court filings also linked him to internal searches. Employees had also raised concerns about shared systems that could expose technical details beyond individual teams.
Staff specifically questioned OpenAI’s use of a single code repository, known as a monorepo, for much of its software. Employees with access could reportedly view details about projects outside their assigned responsibilities, including sensitive research.
Some also feared that documents on GitHub and Slack could reveal important information to determined outsiders. Three sources told Reuters that teams did not always review modified internal testing tools for security weaknesses.
OpenAI disputed parts of the report and said its most sensitive work sits in separate repositories with strict access controls. The company added that its monorepo excludes critical model details, including the numerical values known as model weights.
It also said regular security reviews now cover internal tools under Chief Information Security Officer Dane Stuckey. Spokesperson Elana Widmann said, “Protecting our technology, data, and people is a responsibility we take seriously.”
OpenAI has started shifting resources toward security, according to people familiar with the changes. Senior executive Thibault Sottiaux now leads an initiative focused on protecting company systems. In August, he reassigned around one-quarter of production engineers to improve security infrastructure, one source told Reuters. The company also announced another internal security sprint after Reuters requested comment.
The case highlights growing pressure on AI labs to balance rapid development with stronger protection for valuable research.
Also Read: OpenAI Changes Data Center Terminology as AI Energy Concerns Grow