Reuters: OpenAI employees warned for years about internal security lapses

Some OpenAI employees raised concerns internally about the company's security practices even before ChatGPT launched in late 2022, Reuters reported on Tuesday, November 3, citing five people familiar with the matter.

Years of internal warnings

According to the sources, employees shared their concerns with top leaders and the board of directors, and some came away concluding that management was more focused on keeping pace with rivals such as Google and Anthropic than on security. They argued that controls were lax enough that outsiders, including foreign governments, could obtain sensitive information about OpenAI's technology with relative ease.

One example cited was OpenAI's practice of storing most of its code in a single "monorepo," which meant anyone with code access could see technical details of many projects, including sensitive ones unrelated to their work. Three people also said the company had not regularly reviewed the security of internal tools such as the sandboxes researchers use to stress-test new models.

OpenAI's response

OpenAI spokesperson Elana Widmann said the company is one of the most targeted organizations in the world and invests heavily in protecting against insiders, cybercriminals and nation-states. "Protecting our technology, data, and people is a responsibility we take seriously and a matter of national importance," she said. OpenAI said its most sensitive work sits in separate repositories with strict access controls, that the monorepo does not contain model weights, and that it now conducts regular security reviews of internal tooling.

Two people told Reuters that Thibault Sottiaux, a senior executive overseeing product development, has taken charge of an internal initiative to secure OpenAI's systems and in August reassigned a quarter of production engineers to security infrastructure. After Reuters sought comment, OpenAI announced its next internal security sprint and expanded its bug bounty rewards.

Wider context

Reuters noted that in October the US Justice Department charged a then-OpenAI engineer with planting a hidden recording device in an executive conference room, and that a cybersecurity startup said in late October it had accessed some of Anthropic's internal systems. The report highlights the tension between the open research culture at leading AI labs and the need to protect increasingly strategic technology, an issue that could also complicate OpenAI's and Anthropic's planned public listings.

Sources: Reuters, OpenAI