Meta is now tapping into a novel resource to train its AI systems: the activity of its own employees. The company intends to leverage data gathered from staff members' mouse movements and keystrokes as it works to develop more advanced and efficient artificial intelligence. This approach, initially covered by Reuters, illustrates the extensive measures technology firms are taking to secure fresh training data—essential material that enables AI models to learn how to perform tasks more effectively and answer user inquiries.

EMBED_PLACEHOLDER_0

To facilitate this, Meta is introducing an internal tool designed to record such inputs within specific applications, aiding in model training. The company has implemented protective measures to shield sensitive information, ensuring the data is utilized solely for this purpose.

This development highlights growing privacy concerns within the AI sector, where yesterday's internal corporate discussions are increasingly being repurposed as raw material for a new kind of corporate supply chain. Just last week, reports emerged that defunct startups were being mined for their archived corporate communications—drawn from platforms like Slack, Jira tickets, and other internal messaging systems—which can then be transformed into fuel for AI training.