Meredith Whittaker

In the [YOU] on AI Field Guide

[YOU] on AI is written from inside the exhilaration of capability—the builder discovering that a single engineer with the right tool can do the work of twenty, feeling the ground shift beneath the experience. Whittaker is the voice the cycle most needed to hear at three in the morning when the tool felt like pure liberation. She does not tell anyone to put it down. She asks them to look up—to see the handful of companies whose concentrated power makes the magic possible, and to ask whose freedom expands and whose contracts as a result.

Her lens reframes every question the cycle asks about AI's social consequences. The issue is never whether the capability is real—she does not dispute it—but who owns the infrastructure that delivers it, on what terms, and against whom it can be aimed. The same model that generates code can generate profiles. The same infrastructure that delights the builder surveils the consumer. Her insistence that the Venn diagram of the AI industry and the surveillance industry is a circle is an invitation to see the continuity between the tool that empowers and the tool that watches—to recognize that they are, increasingly, the same tool, owned by the same companies, fed by the same extracted record of our lives.

She stands in the cycle's gallery as the thinker who refuses the artifact's seduction most completely. Where Kate Crawford maps the material costs of AI infrastructure and Shoshana Zuboff names the economic logic behind behavioral extraction, Whittaker adds the insider's authority: she did not theorize about the system from outside. She built it, read its contracts, watched its incentives operate, and then walked away. Her authority is the authority of the defector, and in the cycle it functions as a permanent corrective to the narcotic of capability.

The practical expression of her critique is Signal—an existence proof that communications technology can be built on principles diametrically opposed to those that produce AI as she analyzes it. The cycle that began with the orange pill must also reckon with the anti-pill: the deliberate architecture of refusal, the technology engineered to collect nothing, to encrypt everything, to place private communication structurally beyond the reach of those who would surveil it. Whittaker's career is the demonstration that the critique and the construction are two halves of a single project.

Origin

Whittaker earned a degree in rhetoric and English literature from the University of California, Berkeley—an education in how language structures power, which would prove more useful than any computer science credential for the argument she would spend her career making. She joined Google in 2006, where she founded the company's Open Research group and co-founded Measurement Lab. She was not a critic peering in from the academy. She was an insider with access to the dashboards, the budgets, the quiet meetings where the future is priced and provisioned.

The walkout of 2018 was the hinge of her intellectual life because it collapsed a distinction that most technology workers are trained to maintain: the distinction between the technical and the political. The workers who built the tools had almost no say in whether those tools would be sold to the military or deployed for surveillance—decisions made above them by people answerable to shareholders rather than to the public or even to the engineers themselves. Whittaker drew from that experience a conviction that recurs throughout her work: the questions worth asking about AI are not primarily questions about the technology's capabilities but about the structure of power within which those capabilities are developed and deployed.

In 2017, before the walkout, she had co-founded the AI Now Institute at NYU with researcher Kate Crawford—one of the first research centers dedicated specifically to the social implications of artificial intelligence. Its annual reports documented concrete harms that celebratory coverage obscured, and its insistence on studying power rather than merely bias shaped the emerging discourse on AI ethics. Since September 2022 she has served as president of the Signal Foundation, where the critique she built in text she now builds in code.

Key Ideas

AI is a surveillance technology. In September 2023, Whittaker offered the single sentence that has come to define her public argument: AI is a surveillance technology. She elaborated with a Venn diagram that is a circle—the AI industry and the surveillance industry are not adjacent or overlapping but the same thing seen from two angles. Surveillance is not a side effect of AI; it is its precondition. The large-scale language models require enormous quantities of data harvested by the surveillance-advertising business model that companies like Google perfected over two prior decades. AI is what that business model produces when given enough data and computation to generalize.

Power, not just bias. Whittaker's reframing of the AI ethics discourse cuts beneath the bias framing that dominated early scholarship. A bias framing locates harm in a system's inaccuracy and implies that the underlying system is legitimate and merely needs calibration. Her power framing asks instead who wields the system and against whom. A perfectly accurate facial-recognition system is more dangerous than a flawed one—it is a more effective instrument of the surveillance and control it was built to serve. This logic drives her advocacy for prohibition rather than mere reform of certain technologies: the danger is not that they fail but that they succeed.

The steep cost of capture. Her 2021 essay of that title argues that the field charged with studying and regulating AI is deeply dependent on resources from the very industry it is meant to scrutinize. Because frontier AI requires data, compute, and funding that only the major technology companies can provide, those firms control the conditions under which research is conducted. The effect is not crude censorship but something subtler: the gradual narrowing of acceptable questions. She draws the parallel deliberately to tobacco and fossil-fuel industries, which funded research designed to manufacture doubt. The captured ecosystem produces epistemic capture—a consensus that should be read with the political economy of its production in view.

AI is not magic. Whittaker's most persistent rhetorical intervention: AI is not magic. It is a product of concentrated computational power, concentrated data resources generated through surveillance, and the concentrated power of the companies that control these inputs. The magical framing performs a specific function—it obscures human agency and responsibility, converting choices made by identifiable actors into outputs of an autonomous force. By substituting material description for magical framing, she reattaches responsibility to the people who hold it and restores to ordinary citizens the sense that they have standing to contest the technology.

Signal as architecture of refusal. Signal is engineered to refuse extraction: end-to-end encryption, minimal data collection, nonprofit structure. Where the AI giants are structured to extract and aggregate data, Signal is structured to refuse it. If AI is the apotheosis of the surveillance business model, Signal is its negation—a working demonstration, used by millions, that the negation is possible. Her defense of encryption against government backdoor proposals follows the same logic: any mechanism that allows authorized access also creates a vulnerability that others can exploit. The mathematics does not permit selective surveillance.