Meta installs work tracking software on employee devices to train up AI. Inevitable backlash ensues

Meta is installing work tracking software on the computers of some employees that records mouse movements, keystrokes and other interactions in order to train its artificial intelligence systems, according to reports. The programme, introduced for US-based staff, is designed to capture how people use workplace software in real time, including actions such as selecting menu options, using keyboard shortcuts and navigating applications. The data may also include periodic screenshots which Meta says is to provide context for these interactions.

The company says the aim is to improve the performance of AI tools, particularly systems intended to carry out everyday digital tasks. By analysing how employees complete routine work on computers, Meta is seeking to generate more detailed training data than is typically available from existing sources.

According to reports, the tracking software operates within a defined set of work-related applications and does not extend to employees’ personal devices. The company has also stated that the data collected will be used solely for AI training and not for employee performance assessment.

Coverage from the BBC noted that the initiative forms part of a wider effort across the technology sector to secure higher-quality datasets for AI development, particularly for systems that interact directly with software interfaces. Such data is harder to obtain than text or image datasets, prompting companies to explore new methods of collection.

The introduction of the programme has prompted internal concern among employees. Reports from Business Insider describe reactions on company forums in which staff questioned the lack of an opt-out option and expressed discomfort about the level of monitoring involved.

In one internal discussion, employees asked how to avoid participation, while responses from senior executives indicated that the software would be mandatory on work-issued devices.

Meta has said that safeguards are in place to protect sensitive information and that monitoring of employee activity on company devices is already standard practice to some extent. The new system is presented as an extension of existing policies rather than a separate form of surveillance.

The development follows earlier scrutiny of Meta’s use of personal and user-generated data to train AI models, which has drawn attention from regulators and privacy groups in Europe and the UK. Concerns in those cases have focused on transparency, consent and the legal basis for processing large volumes of data.

Meta has said it is planning to invest up to $135 billion in AI infrastructure this year, almost double the $72 billion it spent in 2025, a year in which the business generated $115.8 billion in revenue.