The week that Meta employees became training data

The buried angle here isn't surveillance itself but the specific mechanism: employees' work product is being repurposed as training data, which means the workforce reduction and the data acquisition are not separate cost-cutting and R&D decisions but the same decision. Workers are being extracted from twice.

WIRED's April 17 piece on AI writing tools in newsrooms ('AI Drafting My Stories? Over My Dead Body') framed the labor question around editorial resistance and productivity tradeoffs, but that story assumed workers retained some negotiating position. Meta's reported approach suggests a harder version of the same dynamic: the output of knowledge workers becomes model training material before those workers are let go, which removes even the theoretical leverage of withholding labor. This connects to the broader pattern visible across our recent coverage of OpenAI shedding teams and Anthropic navigating government relationships: large AI labs are consolidating control over data and talent pipelines simultaneously, and the workforce is increasingly a means to that end rather than a protected input.

Watch whether any Meta employees pursue legal action specifically over the training-data use of their work product in the next 90 days. A filed claim would force a public accounting of whether standard employment agreements actually authorize this practice, and the answer would set a precedent that extends well beyond Meta.