The Surveillance Business Model

In the [YOU] on AI Field Guide

[YOU] on AI describes generative AI tools as instruments of creative liberation—tools that let an individual builder summon working software from a conversation, that collapse the imagination-to-artifact ratio toward zero. Whittaker's surveillance business model analysis does not deny the experience. It reframes the infrastructure on which the experience depends. The same models that generate code can generate profiles. The same infrastructure that delights the builder surveils the consumer. When an AI assistant is embedded in your email, your documents, your calendar, your code, it is not only helping you—it is also, potentially, observing you, and the observation is the price of the help. The surveillance business model, in this analysis, does not retreat as AI advances. It rides AI into every corner of life that AI is invited to enter.

The model also explains why AI capability is so concentrated. The barriers to building frontier systems are not primarily intellectual. The relevant techniques are widely published and broadly understood. The barriers are material—the data, the compute, the capital—and those barriers are precisely what the surveillance business model accumulated for a small number of firms. This is not incidental to the technology. It is inherited from the economic model that birthed it. And it has a crucial consequence for the optimists who imagine that AI might eventually loosen the grip of the surveillance giants: because AI requires the very resources that only the surveillance incumbents possess, it deepens their advantage. The technology that was built to be democratizing is, structurally, a centralizing force.

Shoshana Zuboff's concept of surveillance capitalism named the broader economic system; Whittaker's surveillance business model concept narrows the analysis to the specific resource-accumulation mechanism that explains why AI looks the way it does, concentrates where it concentrates, and serves the interests it serves.

Origin

Whittaker developed the concept across her work at the AI Now Institute, her public testimony, and her 2023 TechCrunch Disrupt keynote, where she stated the core claim in a single sentence: AI is a surveillance technology. She elaborated with the Venn diagram image: if you drew the AI industry and the surveillance industry as two circles, the diagram would be a circle. The two are not adjacent or overlapping. They are the same thing seen from two angles.

The argument builds on the institutional genealogy she traced in her 2023 essay 'Origin Stories: Plantations, Computers, and Industrial Control,' which connects the computational project of managing and measuring labor to the surveillance infrastructure of contemporary AI. The essay argues that the surveillant and controlling character of AI is not a recent corruption but a continuity—that the systems being marketed today as unprecedented intelligence are the latest iteration of a centuries-long project whose orientation toward surveillance was present from the beginning.

Whittaker is careful to distinguish the surveillance business model argument from the privacy argument. The privacy framing treats surveillance as a cost we pay for the benefits of intelligent systems, something that better regulation might one day reduce. Her framing reverses the causality: surveillance is not a side effect of AI but its precondition. This reversal is the argument's cutting edge.

Key Ideas

Two senses of surveillance. AI is surveillant in two distinct senses, and both matter. First, AI is built on surveillance: its training data is the residue of pervasive data collection, and its capabilities scale with the intimacy and comprehensiveness of that collection. Second, the use of AI is itself surveillant: when you walk past a facial-recognition camera, the system produces data about you—about your identity, emotion, character—whether or not those determinations are accurate. The deployment of these systems generates new surveillance even as it consumes the old. The circle feeds itself.

Concentration as inheritance. The surveillance business model explains why the AI industry is so concentrated without invoking either the genius of the founders or the natural monopoly dynamics of network effects. The concentration is straightforwardly material: the inputs that frontier AI requires are the inputs that two decades of surveillance accumulation produced. The barriers are data, compute, and capital—and those barriers are owned by the incumbents. No amount of algorithmic cleverness can substitute for the training data that Google's surveillance of billions of users generated.

Privacy as accelerating casualty. Whittaker has argued that AI poses an enormous privacy risk because it calls for the creation and collection of ever more intimate, pervasive, and invasive data. The logic is structural: if the value of a system scales with the quantity and intimacy of its training data, then the commercial imperative is to collect more, and more personal, data over time. Privacy has been a casualty of the tech business model for a long time. AI does not reverse that trajectory. It accelerates it, because the model's appetite for data has no natural ceiling.