OpenAI engineer allegedly installs surveillance device in executive meeting room

A real-life Silicon Valley eavesdropping drama is unfolding inside OpenAI. An engineer with experience at top tech companies is suspected of installing secret recording devices in executive meeting rooms. Surprisingly, the case involves not only a tiny development board costing less than $14, but also a vast trove of interaction logs with AI chatbots. Ultimately, OpenAI itself leveraged its own AI model to conduct the investigation. This incident has prompted a renewed reflection: when AI has become a core asset for tech companies, where might the real security threats truly originate?

Article author and source: ME News



September 30, 2026, San Francisco.

While conducting an anti-eavesdropping inspection in an executive meeting room, OpenAI’s security team discovered something unusual.

A small electronic device, wrapped in tissue paper, is plugged into the USB port on the back of the conference room TV. It appears unassuming but has audio recording capabilities, with internal data encrypted.

The problem is that this is not an ordinary meeting room.

According to information disclosed by the U.S. Department of Justice, this meeting room was primarily reserved for OpenAI executives and their assistants to discuss matters that may involve unreleased technological development, product planning, business strategies, and sensitive issues with government entities and partners.

In other words, the conversations happening in this room may be more sensitive than many files stored on the company’s servers.

Investigators then turned their attention to an internal employee: 35-year-old Saudi engineer Abdulrahman Alfozan.

He joined OpenAI in July 2025, following prior work experience at technology companies such as Meta. According to media reports, he also holds educational backgrounds from MIT and the University of California, Berkeley.

But why would this engineer be near an executive meeting room he doesn't have access to?

Surveillance footage shows that he appeared multiple times in the relevant area before the device was discovered.

More interestingly, after the recording device was discovered, OpenAI installed cameras in the meeting room. On October 5, the cameras captured Alfozan re-entering the room and inspecting the spot on the back of the TV where the device had originally been installed.

This footage subsequently became one of the pieces of evidence disclosed by the prosecution.

The most striking detail in this case is the device itself.

According to Wired, the investigation documents identified the hardware used as the Seeed XIAO ESP32-S3 Sense, a thumb-sized development board.

This device can be equipped with a digital microphone, camera sensor, and microSD storage functionality, offering strong programmability. The retail reference price for the base development board is only $13.99, though this does not represent the full cost of a complete recording setup.

A tool originally intended for developers and hardware enthusiasts was accused of being used to secretly eavesdrop on executive meetings at tech giants.

OpenAI stated that the device features advanced encryption, audio monitoring, and data transmission mechanisms designed to evade detection.

However, there is still disagreement over just how complex the device is. TechCrunch notes that it remains unclear what significant technical differences it has compared to ordinary recording devices.

More importantly, OpenAI security personnel attempted to extract data from the device's storage card but were unable to view its contents directly due to encryption.

This means it cannot be simply concluded whether the device has successfully recorded the meeting, what was recorded, or whether the related data has been transmitted.

But this incident has revealed a concerning fact: in today’s tech companies, the most sensitive information isn’t always confined to servers.

Sometimes, it may also exist within an ongoing executive discussion.

If the device hidden behind the television has the feel of a classic spy movie, then the next discovery is entirely of the AI era.

According to the criminal complaint released by the U.S. Department of Justice, Alfozan repeatedly used internal company tools to query the office building address, meeting room locations, room capacities, and schedules prior to the device installation.

These queries also involve the locations of executive and external partner meetings.

Meanwhile, investigators found that a chatbot account associated with the individual’s spouse and sharing multiple identifying features with the individual’s own account left numerous inquiries related to recording devices between September 24 and 30.

These topics cover how to set up a continuously operating recording device, adjust wireless network connections, improve real-time monitoring performance, and encrypt storage cards.

More specifically, between September 27 and 28, the account also inquired about technical details directly related to the involved device, including custom interface information and specific MAC addresses.

It should be emphasized that the prosecution has not proven that Alfozan’s spouse personally entered these queries, nor has it publicly confirmed who actually input all the chat content.

From an investigative perspective, these digital records establish noteworthy correlations with devices, times, and office spaces that warrant further investigation.

This has introduced a rather ironic twist to the case:

A suspected eavesdropping incident within a top global AI company left digital traces of interactions with an AI chatbot.

On the other side, OpenAI stated that the company also used its own AI models to analyze the devices and their origins, combined with other evidence, to identify the involved employee.

AI appears in the digital evidence related to the alleged misconduct and is also involved in the company’s investigation of the incident.

Of course, AI did not conduct the entire investigation alone. Surveillance footage, internal access logs, physical evidence, and the work of the security team were equally crucial.

There’s another detail in the case that could almost be written directly into a mystery screenplay.

On October 5, Alfozan left an unattended backpack in a printing room at OpenAI.

After company security personnel inspected the backpack, they found a wallet, identification documents, and another device similar to the meeting room recording equipment.

This means that, at least according to the evidence gathered by investigators, there are two related devices.

According to the criminal complaint cited by Wired, Alfozan later reported the backpack lost to the company and retrieved it from the lost and found.

He then reported to the relevant personnel that some items were still missing from the backpack, including credit cards, approximately $500 in cash, and two AirPods.

However, the report noted that he did not mention the second recording device that had already been removed by security personnel.

This incident has raised more questions about the investigation.

Where was the second device originally intended to be placed? Are there any other devices still in existence? How long did the recording device in the meeting room actually operate?

OpenAI stated that the evidence the company possesses indicates that the involved employee may also have planned to install additional listening devices in other sensitive areas of the office building.

However, it is currently unclear whether these plans were actually implemented or whether other participants were involved.

On October 9, the FBI arrested Alfozan.

On October 13, the U.S. Department of Justice formally filed charges alleging that they illegally intercepted oral communications using electronic, mechanical, or other devices.

If ultimately convicted, he could face up to five years in prison, a $250,000 fine, and supervised release after sentencing.

OpenAI has terminated his employment and stated that there is currently no evidence indicating he unauthorizedly accessed or stole model weights, source code, or customer data.

This is an important distinction.

Allegations of secretly recording executive meetings do not equate to confirmed leakage of core technologies; allegations of eavesdropping cannot be directly equated to espionage activities directed by foreign governments.

To date, prosecutors have not publicly disclosed a clear motive for the crime, nor have they confirmed any involvement by a foreign government. Alfozan’s nationality cannot be used as evidence that his actions were directed by a foreign government.

However, the U.S. government has clearly framed this case within a broader context of technological security. Department of Justice and FBI officials emphasized in their statements that safeguarding advanced AI technologies involves not only commercial interests but also U.S. economic competitiveness and national security.

This is precisely what makes this case noteworthy.

In the past, when people discussed AI company security, they often first thought of hacking, model jailbreaking, data breaches, or stolen model weights.

However, this incident serves as a reminder that advanced cybersecurity systems cannot automatically eliminate real-world security risks.

Internal employees may have access to office environments, where unattended hardware ports in meeting rooms could pose risks, and offline discussions that leave no digital records may contain highly valuable trade secrets.

AI can help businesses analyze anomalous behavior, but it cannot make traditional physical security issues disappear overnight.

Even as AI tools gradually lower the barriers to software development, hardware programming, and information organization, companies must reassess the risks of technological capabilities being misused.

This does not necessarily mean that traditional eavesdropping activities have become widespread, but it does indicate that corporate security is no longer just about protecting servers and code.

On October 19, Alfozan will have his next detention hearing. The FBI’s investigation is ongoing, and his ultimate legal liability will be determined by the court.

For OpenAI, this case raises at least one practical issue:

When a company’s core competitive advantage is built on cutting-edge artificial intelligence, how can it defend against security threats that don’t require breaking into the AI model—only gaining access to a conference room?

Sometimes, protecting the next generation of AI technology may begin with checking a USB port behind your television.