Meta’s New AI Plan: Lay Off a Bunch of Workers, Surveil the Rest
As Mark Zuckerberg frantically pivots to AI, Meta employees are left footing the bill.
“According to an internal memo” is the official Hard Reset phrase of the week, all thanks to Meta leaking like a sieve.
On Tuesday, Reuters scored an exclusive: according to an internal memo, Meta plans on requiring its U.S.-based employees to install tracking software that captures their “mouse movements, clicks and keystrokes.” This draconian and creepy plan is in service of training AI agents “in areas where they struggle to replicate how humans interact with computers, like choosing from dropdown menus and using keyboard shortcuts,” Reuters reported.
The tracking software is reportedly called Model Capability Initiative, or MCI. Additional reporting by CNBC revealed a portion of the sites and platforms that the MCI tool might track on work computers, including Slack, GitHub, Threads, LinkedIn, Wikipedia, and even Google. That is a lot of the internet, and also how people communicate at work! The list is apparently tentative, probably because of media scrutiny and a deeply negative response from Meta workers. From CNBC:
“Multiple Meta employees characterized the data-tracking project as ‘dystopian’ in internal messages viewed by CNBC. Others expressed concerns that MCI could widely expose sensitive data, including user passwords, details about new product development, and personal information about workers’ immigration status, health or family members.”
These are all plausible concerns, and they don’t even get into the other glaring issue with companywide tracking software: it can be used to surveil and then discipline workers. Meta’s half-assed assurances that it won’t spy on anyone read like they came from a winking, fingers-crossed Mark Zuckerberg (who, it should be noted, previously got caught taping over his laptop webcam).
MCI can (allegedly) only read screen contents, not files or attachments, the memo reportedly said. Meta spokesperson Andy Stone told Reuters that MCI “would not be used for performance assessments or any other purpose besides model training and that safeguards were in place to protect ‘sensitive content.”’ He did not explain what the safeguards are, nor did he provide new details to CNBC, which seemed to receive a similar, blanket statement.
According to the internal memo, any employees who aren’t assuaged by “trust me, bro” should take solace in how they “can control what shows up on your screen by not doing personal work on your work computer.” Some light scolding on top of some heavy spying! A nice touch.
On Thursday, Meta presented its workers with another option for avoiding the MCI tool. According to an internal memo obtained by the New York Times, “Meta plans to cut 10 percent of its work force, or roughly 8,000 employees, and close another 6,000 open roles.” The steep cuts have been rumored about for a while, but are no longer a rumor—a company spokesperson confirmed them to the Times.
Here’s the crucial part of the layoffs memo, from Chief People Officer Janelle Gale: “We’re doing this as part of our continued effort to run the company more efficiently and to allow us to offset the other investments we’re making.”
As Times reporter Mike Isaac pointed out, the “other investments” are AI. That’s it. Meta is projecting a massive increase in spending this year compared to last, and it’s not because the company is coming up with a fancy new Quest 4 VR headset.
Meta is hardly the only tech company that’s going all-in on AI while cutting down on staff, but it might be the most brazen. It’s not like Zuckerberg and Co. are on a hot streak of exciting new product offerings. The Metaverse is dead and Facebook is broken/impenetrable. Meta AI Glasses have been a commercial success, but they are running up against dire warnings from advocacy groups that are anticipating all sorts of near-future abuses.
No tech company should be trusted to thoughtfully, responsibly transition to an AI-centric business model. That’s especially true of Meta. Even if there’s absolutely no extracurricular funny business with the MCI tool—no surveillance, no scrutiny, purely just passive AI training—the fact remains that Meta is requiring its workers to knowingly assist in their own demise. Either get laid off now because of mounting AI investments, or stick around until AI can do a vague impression of your job, at which point C-suite executives who barely understand the underlying technology will lay you off via another internal memo with a lazy, recycled message. Lather, rinse, repeat.
It should not be this way, and I’m sorry to the affected Meta workers. If you’re a current or former Meta employee, let’s talk: I’m ajshultz.13 on Signal. Conversations are off-the-record unless/until we both say otherwise.
Here’s what else we’re reading this week:
There’s been limited high-quality polling of California’s governor’s race since frontrunner Eric Swalwell dropped out in disgrace. The early returns show a befuddling bump for Xavier Becerra, who headed the Department of Health and Human Services during the Biden Administration—and crucially, wasn’t very good at his job. Despite the winnowing field of weak candidates, Silicon Valley wunderkind Matt Mahan is still polling around 5%. Kind of impressive.
The Verge’s Hayden Field has a sharp feature about how the “AI free ride is over.” As she notes, “After years of offering cheap or totally free access to advanced AI systems, the bill is starting to come due—and downstream, users are beginning to feel the pinch.” Companies like OpenAI and Anthropic are desperate for revenue before investors get impatient. Can they develop an effective business model before it’s too late? I am skeptical!
Palantir just landed another government contract: $300 million to “modernize” how the Department of Agriculture “delivers services to America’s farmers.” One part of the contract is reportedly focused on updating the digital tools that farmers can use to access USDA services, which is innocent enough, relative to Palantir’s typical endeavors. The other part of the contract is more ominous-sounding, even filtered through a whole bunch of useless jargon. From the press release: “Farm security is national security. Through this partnership, Palantir is empowering USDA with core capabilities that will enable it to secure American farmland, enhance supply chain resilience, and shield agricultural programs from fraud, abuse, and foreign adversary influence.”
A handful of people have already gained unauthorized access to Anthropic’s supposed all-powerful, exclusive “Mythos” model, according to Bloomberg.
U.S. Southern Command announced the establishment of an “Autonomous Warfare Command Center,” which will eventually “be dedicated to employing autonomous, semi-autonomous, and unmanned platforms and systems.”
Relatedly, General Dan Caine spoke at a “global security summit” this week, and said that autonomous weapons will be “a key and essential part of everything that we do.” It seems like we’re already there, but I suppose things can always get worse.