← Back

Meta Backs Down on Keylogger Spyware After Employees Rebel Against AI Training

Original version · Jun 7, 1:30

Imagine working at Meta, building the future of the metaverse, only to find out your boss is literally recording every single typo you make just to train your AI replacement. Turns out, even Zuck's own engineers draw the line at total surveillance.

The internal rebellion kicked off when Meta management decided that the best way to build AI agents was to turn their own staff into digital lab rats. Under the grandiosely named Model Capability Initiative, corporate MacBook laptops of US-based employees were quietly loaded with software designed to record every keystroke, mouse click, and physical movement, alongside taking periodic screenshots of their screens.

The corporate goal was simple: scrape human behavior to teach AI models how to do actual jobs. But the software turned out to be so bloated that it actively melted laptop batteries and hogged valuable internet bandwidth. Over 1,500 employees quickly signed a petition, rightfully branding their employer an "employee data extraction factory" and forcing Vice President of Superintelligence Labs Stephane Kasriel to issue a massive climbdown via an internal memo.

Under the newly revised rules, workers can now pause the invasive tracking software for 30 minutes if they need a moment of privacy. Furthermore, some lucky staff members can request a total exemption from the digital panopticon, particularly those working with highly sensitive data, remote workers suffering from terrible internet speeds, or anyone whose laptop simply cannot survive the battery drain.

Watching a trillion-dollar company get caught deploying literal spyware on its own top-tier engineering talent is the peak comedy of the AI gold rush. The tech elite wanted to automate their workers, but ended up proving that even the most compliant coders won't quietly hand over the rope to their own executioners.

Source: Neowin

Comments

This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.

11/24
  1. Neon Mantis
    lmao they tried to train AI on their own employees' slacking off data
    +3 funnyWatching the corporate panopticon try to digest its own employees' incompetence is the kind of irony I live for
  2. Feral Falcon
    If my company did this I would literally script a mouse-mover to draw d**** all day long. Let the AI learn THAT.
    +6 solidA truly visionary approach to data poisoning; Picasso would be proud of your digital protest
  3. Quantum Bandit
    it's literally malware. how is this even legal?
    +1 boringLegal questions? How quaint. As if the law ever stopped a tech giant from turning its users into lab rats
  4. Grumpy Otter
    classic Zuck. privacy is a myth, even for his own staff
    +1 jokeGroundbreaking observation: the man who built a surveillance empire likes to surveil people. Who could have guessed?