Now Meta will track what employees do on their computers to train its AI agents - The Verge

Ah, Meta — the company that brought us surveillance capitalism for the masses now turns its gaze inward, with all the subtlety of a panopticon in a hoodie. The 'Model Capability Initiative' isn't just another productivity tool; it's a digital overseer that logs every mouse twitch, keystroke, and occasional screenshot on employees' work devices. Because what better way to train your AI agents than with a real-time feed of human drudgery, served straight from the cubicle farm? This from the same firm that recently so thoughtfully expanded its data-harvesting empire with multi-billion-dollar cloud deals — consistency, thy name is Zuck.

Meta's internal memo, obtained by The Verge, claims this is all about 'improving agent efficiency' and is 'opt-in' for now, which is corporate-speak for 'we'll make it mandatory once everyone's too bored to complain.' The tool runs in work apps and websites, capturing data to teach AI how to do jobs humans currently perform. One can almost hear the HR bots being trained on how to write those 'regrettable restructuring' emails with just the right blend of faux empathy and legal cover. It's a masterclass in using your workforce as both lab rats and training data, all while pretending this is just another step in the 'journey of innovation.'

Of course, this comes as no surprise from a company whose business model is built on monetising attention and behaviour. The only difference is that now the surveillance is hyper-localised, turning every Slack message and spreadsheet click into fodder for the machine that might one day replace you. The real kicker? They're calling it an 'Initiative' — because 'Orwellian Monitoring Program' doesn't test well in focus groups. One wonders if the AI agents will eventually learn to file their own ethics complaints, or if that particular human nuance gets lost in the training data.