Buy, Don’t Steal: The Economics of Cognitive Warfare

Abstract

Cognitive warfare runs on personal data, and the United States sells much of that data for pennies. This essay does three things: (1) prices what China paid for American voter files; (2) compares that to the documented cost of covert collection; and (3) proposes five actions that raise an adversary’s cost to buy the ammunition.

China did not buy American voter data to read it. It bought raw material for a targeting engine. Commercial advertising spent two decades proving that a large enough pool of personal data, run through predictive models, delivers a different, tailored message to each individual. Point that machinery at a population’s fears and loyalties instead of its shopping habits, and the result is cognitive warfare: the contest to shape what a people believes and how it decides, made precise by data.

A companion essay in this journal argued that cognitive warfare runs on data, and that cognitive defense therefore begins at the data layer. This essay prices that battlefield. A declassified record that the White House released in July 2026 shows that the raw material cost almost nothing, because most of it was legally for sale.

The Arithmetic of Acquisition

In earlier work on compound warfare economics, I proposed the Cost Asymmetry Ratio, the adversary’s offensive unit cost divided by the defender’s response cost, as a measure of fiscal attrition in modern conflict. A Shahed one-way attack drone costs roughly $35,000, give or take $15,000 depending on whose estimate one credits, and forces a Patriot interceptor that the Congressional Research Service prices at roughly $4 million. The Center for Strategic and International Studies (CSIS) adopted that same $35,000 midpoint as a deliberately conservative figure when analyzing Russian drone strikes. The exchange runs better than 100-to-1 against the defender. In that study, 12 of 14 canonical threat pairings favor the attacker. The declassified materials permit the same analysis in a different domain: intelligence acquisition. The results are worse: the exchange favors the attacker by an even wider margin.

Start with what the data costs to buy. Every US state and the District of Columbia sells or releases its voter file. California caps the fee for its entire file at $100, which for roughly 22 million records works out to about a two-thousandth of a cent per voter. Alabama, at the expensive end, charges one cent per name, or roughly $38,000 for its statewide list. Eleven states charge nothing at all, and according to Election Assistance Commission data compiled in 2020, 24 states charged between $1 and $1,000 for a statewide file. A complete national file assembled directly from the states runs to the low six figures at most.

China paid less than that. The clearest document in the declassified release, a network-defense notice, records a Chinese actor downloading the voter files of six states from US commercial websites in January 2022 and states that the information “is publicly available for download from the commercial websites.” No state fee, no intrusion, no exploit. The marginal acquisition cost was bandwidth and analyst hours.

What Stealing Costs

The covert alternative has a price list, and for once it is public. When the Justice Department indicted 10 Chinese nationals in March 2025 for hacking on behalf of the Chinese government, the charging documents described the contractor i-Soon billing Chinese intelligence services between $10,000 and $75,000 for each email inbox it successfully compromised, with further payments for analyzing the contents. Leaked internal files from the same company, analyzed by SentinelOne in 2024, show data from a foreign ministry fetching roughly $55,000. i-Soon employed over 100 people and projected revenues near $75 million.

Those are the discount rates for soft targets, accounts and networks protected by ordinary commercial security. Hard targets are devices and systems that can only be breached with novel unpublished exploits. The exploit brokers that supply state programs publish their bids: up to $7 million for a full-chain iPhone compromise, $5 million for Android. RAND’s study of zero-day vulnerabilities found a median useful life near seven years, which is the asset an intelligence service burns when an operation is discovered. Above all of it sits a national intelligence budget the Director of National Intelligence discloses only as a top line, $73.4 billion requested for fiscal 2025, with everything below that classified.

Set the routes side by side. Ten thousand dollars per inbox is the cheapest documented rate for stealing personal data one target at a time. One hundred dollars buys the registration record of every voter in California. The asymmetry is three to four orders of magnitude in direct cost, and that is before risk is priced at all. Covert collection carries indictment, attribution, sanctions, and the loss of tools that took years to build. A commercial download carries none.

The defender’s side of the ratio is worse still. Equifax agreed to a settlement of at least $575 million and potentially $700 million for a breach affecting 147 million people. UnitedHealth Group reported that the Change Healthcare attack would cost roughly $2.87 billion in 2024. A breach of the Office of Personnel Management compromised background investigation files on 21.5 million people, including 5.6 million sets of fingerprints, at a national security cost no settlement figure captures. On the kinetic battlefield, an attacker spends a dollar to force thousands. In the data domain, the attacker spends nothing to force billions, or, when the data is legally public, to force no response at all, because there is nothing to remediate and no one to indict.

The Portfolio Is the Tell

The declassified catalog of data held by a People’s Republic of China computer network exploitation actor around 2019 reads as my cost-asymmetry framework predicts. Most of its American entries are cheap or free: voter files, breached credentials circulating in criminal dumps, social media account data, the commodity substrate of identity resolution. A small number of entries fall into a different category: F-15 maintenance and technical manuals, and documents from four cleared defense contractors. Technical data tied to items on the United States Munitions List cannot be bought at any price, which is precisely why it was stolen.

That split is what a rational collector produces. Commodity data is purchased or scraped because covert action against it would be economically irrational. Export-controlled technical data is stolen because no legal market exists. Expensive, risky tradecraft is reserved for the material money cannot buy. An intelligence service that behaves this way is optimizing, and the results of that optimization are reflected in the price column.

This reframes how the July release should be read. The alarming interpretation, that a 2022 download of public records constitutes an election compromise, misprices the event. The economic interpretation is more troubling: the United States has structured its data environment so that the cheapest route to a national targeting file on the American population runs through lawful commerce. The vulnerability is not a firewall. It is a fee schedule.

Why the Cheapest Input Buys the Most Influence

Cognitive warfare is the effort to shape a population’s perceptions, judgments, and decisions. Congress ordered the Department of War to define it in the 2026 National Defense Authorization Act, a task still unfinished months past its deadline, and this journal has carried a sustained debate over whether the term describes anything new. Set the doctrinal argument aside, and the mechanism is one the advertising industry already built. Gather enough data on a person, model what moves them, and serve a message calibrated to that individual. Advertising technology runs this logic billions of times a day to sell products. A state with the same pool of data does it to sell doubt and division, one citizen at a time, and adversaries are already contesting that space with AI-generated content at scale.

The declassified materials indicate that China reasons this way. Beside the voter files sits a plan to exploit American fault lines over race, immigration, and the pandemic response, and to push tailored narratives through social and traditional media. Traditional propaganda involves broadcasting one message to everyone and hoping it lands. A data-driven campaign delivers a different message to each person, measures who responds, and adjusts. The difference is the objective. A retailer wants a purchase. An adversary wants a belief.

Here the implications for scale turn the economics from severe to alarming. A broadcast’s reach is capped by its cost. A data-driven influence campaign’s is not, because once the pool of data exists, the marginal cost of one more tailored message approaches zero. China assembled that pool for pennies. The input to one of the most consequential forms of modern influence, the shaping of a democratic population’s decisions, is the cheapest item on the battlefield.

Inverting the Curve

In the kinetic domain, the fix for a bad exchange ratio is a cheaper defense: a laser shot costing a few dollars against a drone costing tens of thousands. The data domain has an analog, and it is regulatory, not technical.

Executive Order 14117 and the Justice Department’s Data Security Program, in force since April 2025, function as an export control on Americans’ bulk sensitive data, restricting transfers to six countries of concern. The economic effect is the point. The rule raises the adversary’s acquisition cost at the cheapest point on the curve, the legal purchase, and pushes collection back toward the covert route, where cost, risk, and attribution all favor the defender. A dollar of acquisition cost imposed at the point of sale buys more security than a dollar of breach response after the loss.

From Arithmetic to Action: Five Recommendations

Five actions raise the adversary’s cost to acquire the ammunition of cognitive warfare.

First, extend and enforce the bulk-data transfer restrictions. Close the reseller and aggregation seams through which restricted categories leak, and fund the enforcement capacity to detect violations, since an unenforced export control prices at zero.

Second, reprice public records for bulk machine access. This does not require closing them. Transparency for citizens, journalists, campaigns, and researchers can coexist with fees, identity verification, and rate limits calibrated to make wholesale foreign harvesting expensive and visible. A journalist checking a voter roll and a foreign actor downloading 51 jurisdictions are different acts, and the fee schedule should treat them differently.

Third, harden the holders whose data cannot be bought. As the legal route closes, adversary covert spending will concentrate on cleared defense contractors, health systems, and genomic repositories. Security investment should follow that predicted shift rather than trail it.

Fourth, extend acquisition-cost logic to minors’ data. Behavioral and biometric profiles built on children age into profiles of cleared analysts and field-grade officers, which makes children’s data the longest-dated asset in an adversary’s portfolio and the least defended.

Fifth, measure defense by the adversary’s cost to acquire. Program evaluation across data security, foreign investment review, and privacy enforcement should report the acquisition cost imposed on a hostile collector, because that figure, rather than defender spending, predicts adversary behavior. The Committee on Foreign Investment in the United States (CFIUS) and Federal Trade Commission children’s-privacy enforcement already generate the underlying data.

Deterrence here comes from arithmetic, as it does against drones: make the cost to acquire this data exceed the influence it buys. Raising that cost is not only privacy protection, or even counterintelligence. It denies easy access to the ammunition needed for the fight over what Americans believe. Today the price runs in the opposite direction by orders of magnitude, and the adversary reads the price list as easily as we do. The cheapest weapon in the arsenal is the one aimed between our ears.

A Note on Estimates and Sources

Voter file figures are drawn from published state fee schedules; the national-assembly range is an inference, not an audited total. Covert collection prices are drawn from the March 2025 Justice Department indictment and from analysis of leaked contractor files. Remediation figures come from Federal Trade Commission settlement documents and Securities and Exchange Commission filings. The Cost Asymmetry Ratio framework is developed in The Cost Curve as Deterrent, with the simulator and data sources on GitHub.

Disclaimer: The views expressed are the author’s own.