OpenAI said Tuesday that the FBI arrested one of its employees after the company used its own models to investigate a hidden recording device discovered inside one of its offices.
The ChatGPT maker said its models helped investigators understand the device, trace its origins and identify an employee suspected of planting it several days earlier. The company subsequently terminated the employee and referred the case to federal authorities.
The FBI investigation remains ongoing. OpenAI has not disclosed the employee’s identity, the office where the device was found or the type of recording equipment involved.
The company also did not explain which models it used or how they helped establish the device’s origins. It remains unclear whether the models analyzed images, technical components, records or other evidence.
Models trace hidden device
In a statement posted on X, OpenAI said it discovered a sophisticated recording device concealed in one of its offices and used its models to investigate the incident.
“We used our models to understand the device and where it had come from, identifying that a specific employee had planted it a few days earlier,” the company said.
OpenAI said it then contacted the FBI, whose agents and prosecutors moved to make an arrest. The company said it would continue cooperating with the investigation.
The company has not released technical details showing how the models reached their findings or how investigators verified the results. Consequently, the announcement does not establish how much of the investigation was conducted using automated analysis rather than conventional security methods.
OpenAI also said it had found no indication that the employee improperly accessed or removed its model weights, source code or customer information.
Model weights are the numerical parameters that determine how a trained model processes information and produces outputs. They are among the technical assets that companies developing advanced models seek to protect.
The company’s statement does not establish what information, if any, the recording device captured before it was discovered.
Security concerns extend further
OpenAI described the incident as an example of using its own technology to investigate potential security threats.
Its security program is designed to address risks involving employees, cybercriminals and nation-state actors, the company said. The incident also comes amid separate security-related developments involving OpenAI.
On Friday, the company said it had dismissed three researchers following an investigation into alleged violations of policies governing sensitive information.
The researchers, Jasmine Wang, Tomek Korbak and Mikita Balesni, had published a letter describing their dismissals and raising concerns that the company’s actions could create uncertainty among employees and discourage researchers from speaking openly about safety issues.
The two incidents involve different circumstances, and OpenAI has not indicated that they are connected.
Earlier this year, a Texas man was charged in connection with an alleged Molotov cocktail attack at the home of OpenAI CEO Sam Altman and an attempt to set fire to the company’s headquarters.