Edition 016: Data is Power

What the data economy has built is something the information economics tradition spent decades analyzing. Joseph Stiglitz, George Akerlof, and Michael Spence shared the Nobel Prize in 2001[1] for showing that markets with asymmetric information — where one party knows more than the other — do not work the way the textbooks say.[2] They produce systematic inefficiencies, skewed bargaining, and outright market failure. The invisible hand, it turned out, was often invisible because it was not there.[3]‍ ‍

In the classical models Stiglitz and his colleagues built, the information advantage ran in a direction that offered individuals a narrow but meaningful protection. You knew more about your own creditworthiness than the bank deciding whether to lend to you.[4] You knew more about your own health than the insurance company selling you a policy. [5] You knew more about your own ability than the employer deciding what to pay you.[6] The institutions were working with aggregate data —credit market statistics, actuarial tables, labor market averages — and trying to extract information they did not have through credit rationing, screening contracts, and efficiency wages. You held a sliver of private knowledge they could not fully access. That sliver was, in a modest but real way, a form of power.

The data economy ended that. Actually, it did not just end it. It reversed it.

The platform you use every day now knows more about you than you can reliably reconstruct from memory. It has logged ten thousand interactions you forgot about and built a picture of your behavior, your preferences, your vulnerabilities, and your breaking points that is more precise than anything you could assemble yourself. It knows what makes you anxious. It knows what makes you click. It knows the difference between a search you make at two in the afternoon from a work computer and a search you make at two in the morning from your phone. It knows you looked at that item repeatedly before you put it in the cart, and it knows you are not going to abandon it now.[7]‍ ‍

Similarly, the lender knows more about your creditworthiness than you know about your options — your income volatility, your spending patterns, your social graph, modeled in ways you cannot see or audit.[8] The employer knows more about what workers like you will accept than you know about what you are worth — it has run its algorithms across millions of applicants and can predict your reservation wage before you name a number.[9]‍ ‍

***

Consider insurance specifically. The institution of insurance was built on a particular kind of ignorance. In his book The Ascent of Money: A Financial History of the World, Niall Ferguson traced, among other things, the insurance system, which emerged in eighteenth-century Scotland, when two ministers used the new mathematics of probability and demographic statistics to build an early scheme of pooled premiums grounded in actuarial projection.[10]‍ ‍

It used to be that an insurer did not know which member of the pool would get sick, or crash, or lose the house. The insured members did not know either. Because no one knew, the cost of the events was spread across everyone. The healthy member’s premium subsidized the sick member’s claim, just as the careful driver’s premium would subsidize the unlucky one’s accident. You paid a small regular amount in exchange for the protection of the pool against the day the event happened to you. That system required the insurer not to know any individual’s precise risk.

Now, an insurer can predict the individual’s expected loss with precision. Personalized pricing means the high-risk member is no longer pooled with the low-risk member. She is priced separately with a personalized premium reflecting her individual expected loss rather than the average loss of the insureds. What remains is a transaction in which the insurer extracts the expected value of her individual loss plus a margin, which she must pay because the law still makes insurance mandatory. But this is no longer insurance against an unknowable event. It is prepayment, with a markup, for an event the insurer has already predicted.

The data economy gave the insurer that capacity. Your wearable knows your heart rate and how often you exercise. Your credit data shows whether you are under financial stress. The telemetry from your car shows how you actually drive, not how you say you drive. None of this is an actuarial table. This data does something different. It predicts what will happen to you, specifically.

The laws that govern the insurance industry were written for a world in which the insurer could not do that. The separate regulatory framework, the rules about how rates are set, the prohibition on charging different prices to different groups of people, the requirement that everyone buy coverage — all of these were built to protect the shared risk. The shared risk is gone. What the industry is selling now, under the legal protection that was built for insurance, is not insurance.

***

Once the institution knows more about the individual than the individual knows about herself, the rules built for the older arrangement keep running on a world that no longer exists.

For years online, we have clicked on traffic lights and crosswalks to prove we are not robots in various reCAPTCHA challenges. When you do that, you provide free labor, labeling images for Google's computer vision model.[11] That model now feeds the same family of computer vision systems Google has used to develop its autonomous-vehicle program. You were not told, asked, or compensated. Now you will hail a driverless car and pay for the ride. Otherwise, you will be surveilled as one of its camera-clad cars rolls by you. And as a recent episode of Hacks pointed out, no one voted to put driverless cars and their cameras on our streets. Instead, the data you provided built the asset, and the asset is being deployed back at you on the streets you walk on, on terms you were never given the chance to set. Every driverless car on the road is also set of cameras on the road, recording the city and everyone in it, and feeding what it sees back into the same system that was built from what you gave it for free.

The cameras do not stop at the curb. When Nancy Guthrie was taken from her home in Tucson in February 2026, she did not have an active recording subscription for her Nest doorbell camera. By the terms Google set, only about three hours of event history would be available,[12] and the morning she disappeared was already out of reach. Days later, Google engineers recovered the footage anyway, from what the company described as residual data in its backend systems.[13] In this case, the recovery served a public purpose, finding evidence related to a missing mother. But no one agreed to let Google keep this data. Nest users buy a device that promises three hours. Google keeps much more. Why is the company collecting it, and what is it doing with it the rest of the time, when there is no missing person and no FBI request and no public-facing reason to admit the footage exists at all?

And the problem persists even when the company does not itself collect the data. Palantir assembles it. Founded in 2003 with funding from the CIA’s venture capital arm, the company builds the systems that pull together what corporations have already gathered — tax records, biometrics, location, financial history, immigration status — and hands the assembled picture to the federal government. Its clients include the Department of Defense, the IRS, and ICE. Palantir has received more than $900 million in federal contracts since the second Trump administration began,[14] and in mid-2025 the Army awarded it an Enterprise Service Agreement worth up to $10 billion over ten years.[15] No one whose data flows through the system agreed to any of it. The government does not need to conduct a search. It buys the analysis. The constitutional protection that bars the state from going through your life directly is silent on the company the state pays to do it instead.

The same infrastructure that knows everything about some people is, on command, made to forget what it knows about others. On May 19, 2026, the Justice Department added an addendum to its settlement with the president. The IRS is now, in the language of the document, forever barred from pursuing any claim against Donald Trump, his family, or his businesses on any tax return filed before that date.[16] A federal statute makes it unlawful for any executive branch official to request, directly or indirectly, that the IRS terminate an audit of any particular taxpayer. A federal statute makes it unlawful for the President, his staff, and most senior executive branch officials to request, directly or indirectly, that the IRS terminate an audit of any particular taxpayer.[17]The addendum was signed by the Acting Attorney General — the one senior official the statute exempts —who had previously served as the president's personal criminal defense lawyer.

The data still exists. The returns are still in the file. Nothing has been resolved on the merits. What has changed is that one taxpayer’s data has been placed outside the reach of the system that holds it. Every other taxpayer’s data remains inside that reach. The IRS will continue to match wages against returns, to flag discrepancies, to issue audit notices, and to refer cases to the Justice Department. It will do all of this with the same data infrastructure that has now been instructed to look past one set of files. The infrastructure does not change. The rule that governs whose life it can be applied to does.

***

That is what data is, once it has been concentrated. And information, at this scale and this precision, is not just knowledge. It is leverage. Whoever holds it extracts value from whoever does not. But it’s not just a record from which value can be extracted, the data you generated is sold back to you (self-driving cars), the data you bought a contract to limit is kept anyway (Nest), the data corporations gathered about you is handed to the state (Palantir), the data the state holds about you can be selectively erased (Trump). So the asymmetry the section has been describing is not only that the institution knows more about you than you know about yourself. It is that the institution decides, case by case, what to do with what it knows. And the people whose data is exempted from consequence are not the people whose data was taken without consent in the first place.

***

The information advantage accumulated through the data economy does not stay in the market. It becomes political power. The platform that knows everything about you also knows everything about your neighbors, your congressional district, your community’s fears and hopes and breaking points. The data broker that can sell your profile to your landlord, can also sell it to the campaign trying to decide whether to turn you out to vote or keep you at home. The employer using an algorithm to screen your application is using the same data infrastructure as the government using an algorithm to decide whether you are a security risk. What began as a market condition—one party knowing more than another, extracting value from that advantage—has become a governance condition.

The people with the most data have the most power. Not just the power to charge you more for insurance. The power to shape what information reaches you before you make any decision at all, including the decision about who should govern the system that is doing all of this to you.

***

All of this comes at a cost. That edition is next.

***

[1]See “The Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel 2001.” The Nobel Prize, nobelprize.org/prizes/economic-sciences/2001/summary/. Accessed 4 May 2026.

[2]See “The 2001 Prize in Economic Sciences.” The Nobel Prize, nobelprize.org/prizes/economic-sciences/2001/popular-information/. Accessed 4 May 2026.

[3]See Joyce Routson, “Joseph Stiglitz: Today's Market Behavior Shouldn't Be a Surprise.” Stanford Graduate School of Business, gsb.stanford.edu/insights/joseph-stiglitz-todays-market-behavior-shouldnt-be-surprise (quoting Stiglitz: “The invisible hand often seems invisible because it’s not there…Markets are not totally efficient.” Accessed 4 May 2026.

[4]See Stiglitz, Joseph E., and Andrew Weiss. “Credit Rationing in Markets with Imperfect Information.” American Economic Review, vol. 71, no. 3, 1981.

[5]See Rothschild, Michael, and Joseph E. Stiglitz. “Equilibrium in Competitive Insurance Markets: An Essay on the Economics of Imperfect Information.” Quarterly Journal of Economics, vol. 90, no. 4, 1976.

[6]See Spence, Michael. “Job Market Signaling.” Quarterly Journal of Economics, vol. 87, no. 3, 1973; and Shapiro, Carl, and Joseph E. Stiglitz. “Equilibrium Unemployment as a Worker Discipline Device.” American Economic Review, vol. 74, no. 3, 1984.

[7]See Nicholas Thompson, “When Tech Knows You Better Than You Know Yourself,” Wired, October 4, 2018, https://www.wired.com/story/artificial-intelligence-yuval-noah-harari-tristan-harris/. Accessed 19 May 2026; see also Yuval Noah Harari, Nexus: A Brief History of Information Networks from the Stone Age to AI (Random House, 2024).

[8] Greenlining Institute, "FCRA and Find Out: How Online Lenders Use Alternative Credit Scores to Prey on Communities of Color." Greenlining, 17 June 2025, greenlining.org/2025/fcra-and-find-out. Accessed 18 May 2026.

[9] Veena Dubal, "On Algorithmic Wage Discrimination." Columbia Law Review, vol. 123, no. 7, 2023, columbialawreview.org/content/on-algorithmic-wage-discrimination. Accessed 18 May 2026.

[10]See Niall Ferguson, The Ascent of Money: A Financial History of the World (Penguin Press 2008).

[11] TechRadar, “Captcha if you can: how you've been training AI for years without realising it“ (January 12, 2018): https://www.techradar.com/news/captcha-if-you-can-how-youve-been-training-ai-for-years-without-realising-it. Accessed 18 May 2026.

[12]Event and 24/7 video history — Google Nest Help: https://support.google.com/googlenest/answer/16593318. Accessed 18 May 2026 (Google’s documentation states that without a Nest Aware subscription, users get up to three hours of event video history).

[13] See Android Authority, “FBI recovers Nest Cam footage without a subscription, raising privacy concerns” (February 11, 2026): https://www.androidauthority.com/google-nest-doorbell-camera-nancy-guthrie-privacy-concerns-3639806/. Accessed 19 May 2026.

[14] American Immigration Council. “ICE to Use ImmigrationOS by Palantir, a New AI System, to Track Immigrants' Movements.” American Immigration Council, 22 Aug. 2025, www.americanimmigrationcouncil.org/blog/ice-immigrationos-palantir-ai-track-immigrants/. Accessed 19 May 2026.

[15] Elizabeth Dwoskin, “Palantir Gets $10 Billion Contract from U.S. Army.” The Washington Post, 31 July 2025, www.washingtonpost.com/technology/2025/07/31/palantir-army-contract-10bn/. Accessed 19 May 2026.

[16] Ryan J. Reilly, “Justice Department Agrees to Drop Any Pending Tax Claims against Trump as Part of IRS Deal.” NBC News, 19 May 2026, www.nbcnews.com/politics/justice-department/doj-agrees-not-pursue-tax-claims-trump-part-irs-deal-rcna345973. Accessed 19 May 2026.

[17]Prohibition on Executive Branch Influence over Taxpayer Audits and Other Investigations, 26 U.S.C. § 7217 (1998), www.law.cornell.edu/uscode/text/26/7217. Accessed 19 May 2026.