Meta has paused the Model Capability Initiative, a program that recorded employees' mouse movements and keystrokes and used the resulting data to train artificial intelligence models, according to reporting by Engadget.
The initiative was launched to help the company develop more effective AI agents by learning from practical usage examples. From the start it attracted significant criticism from Meta employees over the near-constant monitoring; the company had previously made minimal relaxations to the program's rules in response to those concerns.
The immediate cause of the suspension was not employee pushback but a privacy-related mishap: data collected by the tool became accidentally accessible to all Meta employees. The exposed material reportedly included other employees' conversations, performance metrics and other sensitive information.
Business Insider's reporting highlighted that Meta may not have been as prepared for the privacy implications as it claimed when announcing the program. Meta says there is no evidence so far that employees accessed any data without authorization; nevertheless, the company has halted the Model Capability Initiative while it investigates. It has not released further details about the incident.
Why this matters
- The data involved could include highly personal and work-related information, so unintended access raises substantial privacy and workplace ethics concerns.
- The episode underscores ongoing questions about transparency and lawful use of internal corporate data for AI development.
For now, Meta's immediate response is suspension of the program and an internal investigation; the company has said it will provide more information but has not yet shared specific findings.
