Back to news
regulationOpenAI2026-11-04

Reuters reports OpenAI employees raised security concerns since 2022 before eavesdropping device discovery

Reuters cited five sources saying OpenAI employees raised security concerns before ChatGPT launched in late 2022, with policy changes underway after the October recording device discovery.

In October, OpenAI discovered a covertly installed recording device in an executive meeting room, and several former employees said they were not surprised. Reuters, citing five sources, reported that before ChatGPT launched in late 2022, employees had raised concerns about OpenAI security management with senior leadership and the board.

Some employees concluded that management was more focused on catching up with Google and Anthropic than on security. They described OpenAI security management as lax, noting that any employee with code access could view underlying technical detail across many projects. Two sources said OpenAI has begun adjusting its security posture, with senior product lead Tibo Souto steering stronger internal system protection.

The report noted that OpenAI concentrates most code in a monorepo system whose scale and complexity once deterred some potential attackers, but AI agents now have the retrieval ability to sift through material of that size. OpenAI said many of its most sensitive projects live in separate repositories with tightly controlled access, and that the monorepo contains neither all of its code nor model weights.

OpenAI spokesperson Elena Vedman said the company is among the most frequently attacked institutions globally and has invested heavily in systems to defend against insiders, cybercriminals and state actors. The US Department of Justice announced on October 13 that it had arrested a 35-year-old Saudi national accused of secretly installing a recording device hidden in a tissue box in an OpenAI executive meeting room.

OpenAI安全漏洞窃听设备路透社