AI Card Bao AICARDBAO.COM

Industry News

After finding a recorder inside the toilet paper, a former employee of OpenAI said, “No surprise at all.”

In October, a recording device was discovered hidden inside some toilet paper in an executive meeting room at OpenAI. On October 13, the U.S. Department of Justice announced that the person responsible for this act was 35-year-old Saudi citizen Abdulrahman Al-Fozan, who was an engineer at OpenAI at the time of the incident.

The case itself is bizarre enough, but what’s even more intriguing is what happens afterwards. Many former employees were not surprised when they heard about it.

Five people with knowledge of the matter said that long before ChatGPT was launched at the end of 2022, someone had already raised concerns about security issues with senior management and the board of directors. Some employees felt that management placed more emphasis on catching up with Google and Anthropic, and thus gave less priority to safety considerations.

Here’s a concrete example: Those who have access to the code can easily come across a large amount of low-level technical details that are actually unrelated to their own tasks. The reason for this is that OpenAI has, for a long time, stored most of its code in a single repository.

This potential hazard has become more pronounced today. In the past, due to the huge volume of code, it was practically impossible for people to review everything thoroughly; this acted as a natural form of protection, obscuring certain issues. Now, with the help of intelligent systems, all this information can be organized neatly and clearly, which also enhances the efficiency of those who might be trying to hide something.

The company rejects this claim. Spokesperson Wideman stated that OpenAI is one of the institutions most frequently targeted globally, and significant investments have been made in security measures. The most sensitive projects have been moved to separate storage facilities where access rights are strictly controlled. Since the appointment of a CISO in October 2024, regular security audits of various tools are also conducted.

Contradictions are a reality. The United States currently regards AI as a strategic asset and is trying to prevent the leakage of related technology; however, the stricter the control over access to such technologies, the slower the progress of research and development will be. DePiu, a former researcher who left his job in April, clearly stated that speed cannot be sacrificed in the name of security.

Both OpenAI and Anthropic are preparing for their IPOs, so internal control issues can no longer be merely the subject of internal discussions—they must also be addressed in their prospectuses.

Related