OpenAI employees warned of security gaps for years — NDTV
In the United States, OpenAI employees warned management and the board of directors about shortcomings in internal security even before the launch of ChatGPT in late 2022, NDTV reports, citing five informed sources. Some employees believed after such appeals that management was more focused on competing with Google and Anthropic than on protecting the technology.
Access to internal developments
According to the sources, employees with access to OpenAI’s source code could view technical details of many projects, including those unrelated to their work. The company stored most of its code in one large repository — a monorepository. Employees feared that under such conditions, outside actors, including foreign governments, could relatively easily obtain sensitive information.
OpenAI said that its most sensitive developments are stored in separate repositories with strict access controls. The company also stated that the monorepository does not contain all of its code or critically important information about models, including weight parameters.
The recording device case
On October 13, the U.S. Department of Justice announced the arrest of 35-year-old Saudi Arabian citizen Abdulrahman Alfozan, who was an OpenAI engineer at the time. He was charged with installing a concealed audio-recording device in an executive meeting room; according to a court document, the device was hidden in toilet paper.
More current news is available on the UA.News Telegram channel Telegram.
According to court filings, Alfozan also used an AI tool intended for OpenAI employees to find internal documentation about executives’ trips to Saudi Arabia and officials who might attend meetings with Saudi ministers. His lawyers did not respond to a request for comment.
Strengthening protection
According to two sources, senior product development executive Thibault Sottiaux led an internal initiative to protect OpenAI’s systems. In August, he reassigned a quarter of engineers in the production division to improve security infrastructure, according to one source.
OpenAI said it now regularly reviews the security of its internal tools. The company said it had strengthened protections against attacks by insiders, cybercriminals, and state entities, and was preparing the next internal stage of strengthening cybersecurity.
