OpenAI Reportedly Prioritized Product Development Over Safety Controls Long Before Saudi Employee Bugging Incident

Deep News11-04

When OpenAI revealed last month that a listening device had been discovered in an executive meeting room, the reaction among multiple former employees was uniform 鈥?they were not surprised.

According to multiple people familiar with the matter, concerns about the company's safety practices were raised internally well before OpenAI launched ChatGPT in late 2022. These people said employees had expressed such concerns to OpenAI's senior leadership and board of directors; some employees came away from discussions with management believing that leadership was more focused on keeping pace with competitors like Google and Anthropic than on safety issues. The people said these employees believed OpenAI's security controls were very loose, allowing outsiders 鈥?including foreign governments 鈥?to obtain sensitive information about the AI lab's technology with relative ease. For instance, anyone at OpenAI with access to its code could see technical details behind many company projects.

According to two people familiar with the matter, OpenAI has begun adjusting its strategy to address security concerns. They said Thibault Sottiaux, an executive in charge of product development, recently took over an internal initiative aimed at strengthening the security of OpenAI's internal systems. One of the people said that in August, Sottiaux 鈥?often referred to as "Tibo" 鈥?reassigned a quarter of OpenAI's production engineers to the initiative to improve the company's security infrastructure.

However, adding security layers could also slow the pace of AI development and may introduce complications for the large-scale initial public offering (IPO) plans of OpenAI and Anthropic; if those plans proceed smoothly, they could rank among the world's most valuable companies. Will DePue, a former OpenAI researcher who left the company in April, posted on X that with competition so intense, AI industry leaders "cannot sacrifice speed for safety. For them, moving fast matters more than ensuring safety."

When asked about employees' concerns and their view that management did not act on those issues, OpenAI spokesperson Elana Widmann said OpenAI is one of the most frequently targeted organizations in the world. As a result, the company has invested substantial resources in security programs aimed at defending against attacks from insiders, cybercriminals, and nation-state actors. Widmann said: "Protecting our technology, data, and people is a responsibility we take extremely seriously, and it is also a matter of national interest. Our recent actions demonstrate that this security system is working as intended: identifying risks and responding decisively. As technology evolves and threats change, we will continue to strengthen our defenses."

On October 13 local time, the U.S. Department of Justice said federal agents arrested a Saudi national accused of installing a hidden recording device 鈥?disguised inside a toilet paper roll 鈥?in an executive meeting room. The individual, 35-year-old Abdulrahman Alfozan, was an engineer at OpenAI at the time of the incident. According to a court filing, Alfozan also used an AI tool available to OpenAI employees to search for internal documents about executives' visits to Saudi Arabia and to identify which OpenAI senior leaders were most likely to meet with Saudi ministerial officials. Alfozan maintained close ties with a member of the Saudi royal family who had tried to recruit him for a new AI project launched in Saudi Arabia.

According to multiple people familiar with OpenAI's internal operations, some employees have long been concerned about the overly broad openness of information inside OpenAI. At OpenAI, anyone who can access the company's code can view details behind many of its projects, even if those projects fall into sensitive categories and have nothing to do with that employee's work. According to people familiar with the matter, this is because OpenAI stores the vast majority of its code in the same code repository 鈥?a structure known in the industry as a "monorepo." This repository is extremely large and complex, which may have deterred some potential attackers in the past; but today, AI agents are capable of comprehensively combing through and analyzing such massive amounts of data.

People familiar with the matter said that in the past, OpenAI kept most research documents on GitHub and Slack, which raised concerns among some employees that proactive actors could still find key information. But one of the sources and OpenAI said the company restricted access to sensitive work. OpenAI emphasized that most of its most sensitive work is kept in separate code repositories with strict access controls. The company said the "monorepo" does not contain all of the code or key details of its models, including the numerical values that determine model behavior 鈥?the model weights.

According to three people familiar with the company's operations, another issue was that OpenAI previously did not conduct regular security reviews of its internal tools, such as the so-called "sandbox" environments researchers use to stress-test new models. These three people said employees often modified these tools in their work, but those changes were not necessarily security-tested, potentially exposing them to external attacks. The company said that under the leadership of Chief Information Security Officer Dane Stuckey (who joined in October 2024), it has adopted strong security safeguards. OpenAI said it has now begun conducting regular security reviews of its internal tools.

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Comments

We need your insight to fill this gap
Leave a comment