Nine bossware apps, 121 exports of worker names: monitoring is now a vendor-diligence problem

The diligence question at purchase was whether the tool measured what the vendor said it measured. On the evidence of the Vanderbilt, Northeastern and Berkeley test, it does, and it also does something nobody scoped: across nine platforms the researchers logged 121 separate instances of identifying worker data going to outside firms, an average of about 13 per platform, plus 145 third-party domains receiving activity data, roughly 16 per app ^{[1][3][4][14][15]}.

Worker data behaves differently from consumer data once it leaves the building. A consumer ad profile is one person of unremarkable provenance. A monitoring platform's user base is one employer's staffing roster, so a name and a work email travelling together carry the employment relationship with them, alongside whatever the same app reported about IP address, device and browsing ^[3][4]. The investigation's broader finding is that the most common of these programs sent names, emails and other personal worker data to hundreds of data brokers and technology companies without clearly disclosing that they were doing it ^[7]. The employer signed for that. The worker did not, and had no basis on which to ask.

Two of the permission findings deserve separating from the export numbers. A third of the platforms could fix a worker's precise location with the app running in the background or the worker off the clock ^[5]. Three of the nine demanded access to a phone's motion sensors before they would register a shift start ^[6], the same one-in-three proportion ^[17]. Clocking in is a timestamp. Motion sensor access is a description of how a body moves, and whatever the intended use, the collection outlives the shift it was justified by.

The productivity case for the stack was never audited as hard as the workers were. Lannie Duong, a pharmacist at a medical clinic, told the Associated Press her employer timed her phone calls and appointments and queried the length of them in performance reviews; she says she was terminated after taking medical leave ^[12]. Wilneida Negron of Coworker argues that advances in AI and data science let employers assemble dossiers used to punish workers or predict their behaviour ^[9], and that the workers with the least labour market power are where the more intrusive collection gets tested first ^[10]. The reported examples run that way: administrators reading an adjunct professor's written comments to students ^[13], conveyor belt scanners holding warehouse staff to an inspection rate of hundreds of items an hour ^[18].

What a buyer would need to answer a worker's question here is a current data flow list from the vendor, naming recipients. That is a different artefact from a security questionnaire, and on this evidence it is not what anyone received. The per-seat licence was the visible cost. The rest of it sits in the same line item, unpriced, and it was found by academics posing as a customer rather than by any of the customers ^[1].