Ten days separate Meta's announced May 20 layoff from Sunday. The layoff totals eight thousand workers cut plus six thousand cancelled open requisitions, for fourteen thousand affected positions. The rollout was disclosed two days after Meta unveiled its Model Capability Initiative, the internal program that captures employee keystrokes, mouse movements, and on-screen activity across hundreds of websites for AI agent training. The chronology is the story. [1]

The paper's Friday account of the layoff clock running beside the AMD warrant framed the calendar collision. Today the human collision sits in front of it. Meta's Chief Technology Officer Andrew Bosworth, asked in an internal channel whether workers could refuse the MCI logging on a corporate laptop, said there is no opt-out. Reuters confirmed the rollout architecture: every U.S. employee on a work-provided machine is in scope. European employees are exempt because the General Data Protection Regulation makes that kind of monitoring unlawful without explicit consent. [2]

The structural reading is straightforward. A subset of Meta's U.S. workforce is being asked to generate the training data that will be used to build the AI agents that will, by Meta's own product roadmap, do their work. The fourteen-thousand-position layoff, ten days out, is the same workforce delivering the data. Platformer's reading framed the week as one in which "Meta employees became training data," and the timing makes that reading hard to refuse. [3]

What May 20 will produce is the larger institutional question — whether the U.S. legal architecture has any analogue to the GDPR exemption, and whether the cohort of laid-off workers acquires standing to sue over the use of their MCI-captured behavior in models that will outlast their employment.

-- ANNA WEBER, Berlin