Forgot your password?
typodupeerror
Facebook AI Slashdot.org

Meta To Start Capturing Employee Mouse Movements, Keystrokes For AI Training Data (reuters.com) 43

Reuters reports that Meta plans to start collecting U.S.-based employees' mouse movements, clicks, keystrokes, and occasional screen snapshots to train AI agents that can better learn how humans use computers. The tool, called Model Capability Initiative (MCI), will reportedly "not be used for performance assessments or any other purpose besides model training and that safeguards were in place to protect 'sensitive content.'" From the report: Meta CTO Andrew Bosworth told employees in a separate memo shared on Monday that the company would step up internal data collection as part of those "AI for Work" efforts, now re-branded as Agent Transformation Accelerator (ATA). "The vision we are building towards is one where our agents primarily do the work and our role is to direct, review and help them improve," Bosworth said. The aim, he added, was for agents to "automatically see where we felt the need to intervene so they can be better next time." Bosworth did not explicitly spell out how those agents would be trained, but said Meta would be "rigorous" about "building up data and evals for all the types of interactions we have as we go about our work."

Meta spokesperson Andy Stone acknowledged that the MCI data would be among the inputs. [...] "If we're building agents to help people complete everyday tasks using computers, our models need real examples of how people "actually use them -- things like mouse movements, clicking buttons, and navigating dropdown menus," said Stone.

Meta To Start Capturing Employee Mouse Movements, Keystrokes For AI Training Data

Comments Filter:
  • Right. (Score:5, Insightful)

    by Puls4r ( 724907 ) on Tuesday April 21, 2026 @03:04PM (#66105478)
    It won't be used to monitor employees. Until it is. Why on earth would anyone believe what these assholes say?
    • It won't! Once the model is advanced enough, it will simply outperform the employee no performance review needed. Cheers!
    • Re:Right. (Score:5, Insightful)

      by 93 Escort Wagon ( 326346 ) on Tuesday April 21, 2026 @03:22PM (#66105520)

      These people are already employed by Meta. They know what to expect... they've been enabling it themselves, after all.

    • Re:Right. (Score:5, Insightful)

      by goldspider ( 445116 ) on Tuesday April 21, 2026 @03:41PM (#66105538) Homepage

      Why would anyone want to work for a company that does stuff like this??

      • Three reasons (Score:3, Informative)

        1) Food
        2) Shelter
        3) Medicine

        • by Narcocide ( 102829 ) on Tuesday April 21, 2026 @03:53PM (#66105562) Homepage

          4) Continued admission to the Satanic orgies

        • Right, but presumably most of Meta's employees have skill sets that would let them acquire all three at another (less awful) company. This isn't about the bottom of Maslow's hierarchy of needs pyramid [wikipedia.org], it's about the top.

          Meta employees either have no morals, or they believe Meta offers more money and/or status than a job elsewhere ... but they're certainly not there because they otherwise can't afford food or medicine!

          • by taustin ( 171655 )

            Right, but presumably most of Meta's employees have skill sets that would let them acquire all three at another (less awful) company.

            The number of available jobs at less awful companies is probably orders of magnitude lower than the number of employees who would like to leave Meta. Especially considering that the turnover at such companies is most certainly a lot lower.

          • Right, but presumably most of Meta's employees have skill sets that would let them acquire all three at another (less awful) company.

            Not for anywhere near the same Comp&Ben scale.
            Look around. See the increasing pace of convergence and conglomeration? Less awful companies can't compete. A mission statement of "Don't be evil" only carries you so far.

            The real kicker of the absurd "Coal miners can learn to code!" mantra is that we now see it fails both ways -- it's just as absurd to expect a Senior $IT_Job_Title to get laid off and learn to be a plumber crawling through people's attics to connect their new water heater system at age 54.

            • by havana9 ( 101033 )
              Less awful companies can't compete because awful companies are enabled to do awful things, because they are buddies with the legislators they meet in some island where they made another kind of awful things.
              Conglomeration could happen because anti monopoly laws were lifted in the '80s. If in the USA one need basically to have a work to have health coverage it's because insurance companies are thriving on it and are bribing the lawmakers.
    • I'm sure people who work there would know better by now seeing what the company has done to its users?

    • I'd prefer if it was more around content moderation rather than employee value moderation.

    • Soon Meta will release a new LLM for mouse movements that will only be able to jiggle the mouse for hours on end.

      NTTAWWT
    • Did they even think to omit passwords somehow, or will we be able to ask these models in the future: "Give me some random characters that look like passwords, that you might have seen in the past?"

  • by Pseudonymous Powers ( 4097097 ) on Tuesday April 21, 2026 @03:07PM (#66105482)
    Ah, Facebook! Its first generation of employees were quislings for stalking. Its second generation of employees were quislings for advertising. At last, the third generation of employees has arrived: quislings against humanity itself. This will probably be the final generation. Well, at least the final one we humans will be around to see.
  • by drnb ( 2434720 ) on Tuesday April 21, 2026 @03:11PM (#66105490)

    Meta plans to start collecting U.S.-based employees' mouse movements, clicks, keystrokes, and occasional screen snapshots ...

    And when do they start capturing eye movements? They surely want to collect as many "performance metrics" as they can.

    ... will reportedly "not be used for performance assessments or any other purpose besides model training and that safeguards were in place to protect 'sensitive content.'"

    LOL. That's believable, especially at Meta.

    And besides, isn't the goal of the model training to have AI agents "move fast and break things"?

    • by rta ( 559125 )

      They probably mean "sensitive" to the company's interests, not to the employees or users.

      Eye tracking? the EEG caps are probably going to be shipped in soon. All optional. You know, for employees who want to opt in.

      (and/or remain employed for a few more years).

      Oh, don't forget to register for the neural implants beta program announcement list to see when opportunities for more volunteers are available!

      -love,
        Zuck and your HR team

  • by rsilvergun ( 571051 ) on Tuesday April 21, 2026 @03:13PM (#66105496)
    AI training data was going to be the major problem and sticking point. Previously they could just soak it up for free from the internet but now that the internet is 70% AI slop that's not really going to work.

    It does still raise the question of how the name of hell are we going to train AI to do programming tasks when we've replaced most of the programmers. But I guess we will cross that bridge when we come to it. I suspect that over time programming languages will be built AI first and programmer second.

    One thing is for sure everything is going to keep getting worse and worse and worse. And we are going to keep blaming the wrong people because that's what we've always done.
  • "Sir, the system isn't reporting correctly. The data says that our employees only use the left hand for the mouse when surfing incognito."

  • by Voyager529 ( 1363959 ) <voyager529@@@yahoo...com> on Tuesday April 21, 2026 @03:17PM (#66105506)

    our models need real examples of how people "actually use them -- things like mouse movements, clicking buttons, and navigating dropdown menus," said Stone.

    This is the quiet part Stone is saying out loud - the point is to alter UIs.

    Now, the data itself is *probably* helpful...but I am hard pressed to think of ANY application - desktop, mobile, or web - that ANY user would describe as having improved over the past decade. From the disappearance of colors and contrast and borders and scroll bars, to 'settings' screens getting their options eliminated, to toolbar buttons losing their text labels, to modal dialogs and overlays and "hints and tips" taking the place of pop-up ads everywhere...there is VERY little software that has gotten better, despite decades of traditional feedback from users.

    Meta is absolutely going to use this to ascertain how users have figured out to work around the dark patterns and user-hostile design users have spent the past two decades battling, and making it even more difficult and exhausting to get anything done.

    Even if I bought that employees wouldn't be penalized for what the brass finds after putting North-Korean-grade spyware on their computers, there is zero indicating that Meta will be using this to improve anyone's user experience in a way that the user would agree is, in fact an improvement.

  • by Himmy32 ( 650060 ) on Tuesday April 21, 2026 @03:22PM (#66105518)

    Unsurprising that they want data like this when their business model has always been monetizing behavior. But what is interesting is that they didn't just buy some data from monitoring company like Teramind, ActivTrak, a data broker like Thompson Reuters or Lexis Nexis, or some previously captured "Observability" / "RUM" data from an APM company like AppDynamics, DataDog, Dynatrace, or Splunk. But uncategorized data like this tool and those other sources seems less valuable than what could be gotten from a Mechanical Turk or one of the "rewards" softwares/extensions where the action data could be tagged against the intent.

    Maybe I'm too jaded, but I have a sneaking suspicion that the real value is annoying even more employees into leaving.

  • where nobody knows how to do anything, and thus doesn't know if the AI is right or wrong. If someone doesn't know, how can they "guide the AI"? I, for one, like knowing stuff. I like using my brain. I like working stuff out. And that helps me to understand stuff better. With AIs, everyone is going to be stupid.
    • That's the scary part, and it's not talked about enough. AI is helping developers code, but developers are getting worse at coding from scratch because of AI, while juniors never get a chance to find a job because AI can replace them. That's bad enough, but what about new technologies? Who's going to develop new languages, and will AI be able to spit out code using a new syntax? And if not... will there be enough developers left that can build up a codebase in that language for AI to learn? Feels like AI wi
  • "If we're building agents to help people complete everyday tasks using computers, our models need real examples of how people "actually use them -- things like mouse movements, clicking buttons, and navigating dropdown menus," said Stone."

    Why on earth would the computer need the input and visuals that meatbags use ?!?! The mouse clicks and drop down menus are for users that have eyes. The program would not use them.

    • That's where MCP comes in. Everything will be a MCP-connected tool. UI only needed to display lolcats to luser
    • I see a lot of the comments miss the point. The data LLM's were trained on has been exhausted. Everything hukans said or wrote has been "indexed". LLMs can already replace some parts of some jobs but humans stil do things outside of coding and replying to emails with only 5% added value to the existing email thread. In order to replace more humans in more jobs you need AI to learn what people are doing when they do less measurable work: taking decisions, reasoning from disparate sources of information, con
  • Q. CEO: How is the Reverse Centaur project going?

    Staff: We're almost complete, but HR is having some difficulty getting a few employees to sign the employee monitoring arbitration agreement.

    Q. CEO: Who are the employees in particular and what function do they serve in the organization?

    Staff: They are all key employees in our AI department.

    CEO: Hmmmm. How easily can they be replaced?

    Staff: I could take a few months to find employees with a similar background to replace them.

    CEO: What would happen if we just

  • by fahrbot-bot ( 874524 ) on Tuesday April 21, 2026 @03:53PM (#66105566)

    Meta To Start Capturing Employee Mouse Movements, Keystrokes For AI Training Data

    Training your replacement.

  • There are CAPTCHA systems out there on the web that monitor mouse movements to distinguish whether a user is a human or a bot, so as to keep AI web-scrapers out.

    My first thought when reading this, was that perhaps this training on human mouse movements is intended to make AI scrapers that could circumvent those countermeasures ...

  • by dskoll ( 99328 ) on Tuesday April 21, 2026 @04:17PM (#66105604) Homepage

    The answer is Linux. Raspberry Pi Zeros can act as USB devices; it's pretty easy to program one to look like a mouse to the host computer.

    So when you take a break, you plug in your fake mouse that does all kinds of random mouse movements and clicks. Just make sure your screen is locked so none of the clicks do anything.

    • by Himmy32 ( 650060 )
      Then even they weren't going to fire you over performance data, now they will with attempted poisoning the data set. A mouse jiggler may trick your boss looking at Slack/Teams status, but even before AI shenanigans employee monitoring tools were able to catch that.
    • Make sure, related to the Intermittent screen capturing, to play a video on that lock screen, preferably something like the end scenes of The Terminator. Just to send a message what a lowly employee with a mouse is capable of.
  • about:config
    JavaScript disable

    ... and then it all went dark.

  • Corporate America's race to replace humans with AI is going to backlash. Why engage with gamey agents, when you can deploy your own, and wait for the desired result?

    This process is going to repeat, like the Europeans arriving in the Americas, until all the humans are gone, and there's nothing left but bots that do an increasingly good job of acting like us. There will be little reservations, see the Fediverse for an example, where actual humans congregate. There will not be corporate friendly global flat sp

  • It looks like Meta is harvesting training data on how people physically use computers. Technically, I can understand why they would want to do this. And in principle it may result in some AI-assisted time savings for computer users. AI could possibly streamline some mundane browsing and form-filling tasks for you. Some of this typing and clicking stuff could be less necessary. On the dark side, it could push us humans a little farther towards the edge of irrelevance. Eventually we'll just subvocalize our pr

  • If only the staff had union representation

  • Mouse jigglers have been around for years (and became more popular during covid work-from-home conditions)
    e.g https://www.ebay.com.au/itm/31... [ebay.com.au]

    Of course, any competent Meta employee should be able to roll their own in an afternoon.

Asynchronous inputs are at the root of our race problems. -- D. Winker and F. Prosser

Working...