Five people tell Reuters OpenAI staff warned of security lapses for years

Employees at OpenAI raised concerns about the company’s security practices with top leaders and its board, in some cases before ChatGPT launched in late 2022, according to five people familiar with the matter, Reuters reported Tuesday, Nov. 3. The story is by Deepa Seetharaman and was also published by Investing.com and U.S. News.

The five people told Reuters that some staffers concluded management was more focused on keeping pace with Google and Anthropic than on security. Workers argued that OpenAI’s controls were lax enough that outsiders, including foreign governments, could obtain sensitive information about its technology with relative ease, the people said.

The same five people told Reuters that OpenAI stored most of its code in a single repository known as a monorepo, so anybody who could access the code could see details of many projects, including sensitive ones unrelated to their work. They also said OpenAI has kept much of its research documentation on GitHub and Slack. Three people familiar with the company’s practices told Reuters that OpenAI had not done regular security reviews of internal tools such as the sandboxes where researchers stress-test new models.

OpenAI told Reuters that much of its most sensitive work sits in separate repositories with strict access controls, and that the monorepo does not contain all of its code or critical details of its models, including model weights. OpenAI and one of the five sources said access to sensitive work has been restricted, and OpenAI said it now conducts regular security reviews of its internal tooling.

Two people told Reuters that OpenAI has begun to shift strategy and that Thibault Sottiaux, a senior executive overseeing product development, recently took charge of an initiative to secure its internal systems. One of the two said he reassigned a quarter of OpenAI’s “production engineers” to improving security infrastructure in August. That detail rests on one source.

After Reuters sought comment, OpenAI said in a blog post on Monday, Nov. 2, that it is planning its next internal security sprint. The post says the last sprint dedicated 25% of its “product engineers” to finding and fixing weaknesses. It does not mention Sottiaux or August, and the sources reviewed do not say whether the post and the one source describe the same effort.

When asked about the employees’ concerns and their view that management did not act, OpenAI spokesperson Elana Widmann said the company is one of the most targeted organizations in the world and invests heavily in its security program. “Protecting our technology, data, and people is a responsibility we take seriously and a matter of national importance,” Widmann said. “The actions we’ve taken recently show that program doing its job: identifying risks and responding decisively. We’ll keep strengthening our defenses as technology and threats evolve.”

Reuters said the Justice Department announced on Oct. 13 that federal agents arrested Abdulrahman Alfozan, 35, who was an OpenAI engineer at the time, and charged him with installing a hidden audio recording device in an executive conference room. Reuters said his lawyers did not respond to a request for comment. AI Tech Daily reported on Oct. 22, citing CNBC and the BBC, that a judge ordered him held without bail on Oct. 19. He has not been convicted. Reuters also said adding security layers can delay AI progress and possibly complicate OpenAI’s and Anthropic’s plans for initial public offerings.

Five people tell Reuters OpenAI staff warned of security lapses for years

Sources