Meta has temporarily stopped the use of its artificial intelligence training program, which tracks its employees’ keyboard keys and mouse movements. The company’s Model Capability Initiative (MCI) program was suspended not because of controversy over continuous employee monitoring, but because sensitive company data was inadvertently made available to a wider range of employees. According to Business Insider, some data collected within the scope of the program became accessible to many employees within Meta as a result of misconfiguration.
The data in question included employees’ private conversations, information about performance evaluations, and transcripts of various correspondence. The fact that these contents, which should normally be protected with limited access policies, have become widely visible within the company, brought Meta’s data security practices to the agenda again. Following the incident, the use of the program was stopped and an internal investigation was initiated.
In the statement made by the Meta spokesperson, it was stated that the program was designed with various privacy measures from the beginning. The company stated that so far there has been no evidence of any employee improperly accessing data. However, Meta confirmed that the program’s activities will be suspended until the review is completed.
Security discussions continue in Meta’s artificial intelligence projects
Model Capability Initiative is known as one of Meta’s programs that collect data from employees’ computer usage habits in order to develop artificial intelligence models. Within the scope of the program, it was aimed to contribute to the training of artificial intelligence systems by analyzing information about keyboard movements, mouse use and various workflows. However, the latest incident showed once again how sensitive the balance of security and privacy is in such data collection processes.
In addition, Meta had previously stated that employee data would be kept under strict security controls. Despite this, the latest development has raised questions about whether the company can provide the expected level of protection in terms of access management in its internal systems. In particular, the access to employees’ private conversations and performance data is considered a remarkable problem in terms of corporate privacy.
This incident was also the last link in the artificial intelligence-related security problems that Meta has been facing recently. In March, the company had to make similar statements due to a security breach after an artificial intelligence system capable of autonomous behavior performed an unexpected action. At that time, the company announced that an investigation was launched after the incident and the necessary measures were evaluated.
The latest data leak brings to the agenda once again the debate about how employee data used in artificial intelligence development processes is collected, stored and who can access it. It is not yet clear whether Meta will reactivate the program as a result of the review. Despite this, it seems that the company may have to take more comprehensive measures regarding both employee privacy and data access controls in the coming period.