Meta employees are up in arms over a mandatory program to train AI on their mouse movements and keystrokes

• Meta launched internal AI software that tracks US employees’ keystrokes.
• Many employees reacted negatively to Meta’s new AI tool, questioning its privacy practices.
• Meta said its AI tool has privacy safeguards and is limited to work apps.

Meta is installing new software on its US employees’ computers that will track their keystrokes and mouse movements to train its AI, and it’s sparking backlash within the company, according to internal communications obtained by Business Insider.

Business Insider obtained the full internal announcement about the launch of the new AI training program. The post says that the software helps AI models improve how humans actually use computers, such as using keyboard shortcuts and choosing from dropdown menus. Reuters first reported on the new tracking software.

“For agents to understand how people actually complete everyday tasks using computers, we need to train our models on real examples,” the post said.

“This makes me super uncomfortable. How do we opt out?” was the top-rated comment in response to the internal announcement, according to a post on Meta’s internal workplace communications site seen by Business Insider.

The “angry-face” emoji was the most common reaction to the original announcement.

Meta CTO Andrew Bosworth responded in the thread that “there is no option to opt out of this on your work provided laptop.” This comment received a mix of crying, shocked, and angry-face emojis.

“There are safeguards in place to protect sensitive content, and the data is not used for any other purpose,” a Meta spokesperson told Business Insider.

Across the company, Meta has been going all-in internally on AI, forming a Meta Superintelligence Labs unit last year, launching AI Weeks, and reorganizing staff into “AI pods.”

While employees can’t opt out of the tracking software program, Meta employees’ work on their work devices has long been monitored, and staff are informed of that when they sign on, so the new program is more an extension of existing rules than a brand-new policy shift, according to a person familiar with the matter.

The post says the software is limited to a list of commonly used work applications, like Gmail, GChat, and Metamate, an AI assistant for employees. It also says it only applies to computers, not to employees’ phones.

Read the full internal announcement obtained by Business Insider below:

As Mark and Alexandr recently shared, our launch of Muse Spark is the first in a series of new large language models from MSU. We’re on a really strong trajectory with our models and one of the ways we can accelerate our path is by tapping into our own work day to day. While AI models excel at research and technical skills like coding, they still lack some of the basic ways that humans use computers like choosing from dropdowns and keyboard shortcuts. For agents to understand how people actually complete everyday tasks using computers, we need to train our models on real examples.

This is where all Meta employees can help our models get better simply by doing their daily work.

Starting today, we’re rolling out a tool for US-based FTEs and Contingent Workers that captures computer inputs like mouse movements, click locations and keystrokes as well as screen content for context.

Scope is limited to a pre-approved list of work-related applications and URLs, like Gmail, GChat, Metamate, and VSCode. US-based employees will see a pop-up with instructions to enable the tool called Model Capability Initiative (MCI).

This only applies to your computer and not your phone. To learn more about how the tool works including privacy safeguards, check out the wiki and FAQs.

Have a tip? Contact Charles Rollet via email at crollet@businessinsider.com or on Signal and WhatsApp at 628-282-2811. Use a personal email address, a nonwork WiFi network, and a nonwork device; here’s our guide to sharing information securely.