On the 29th of September, 1999, a four-page document was filed with the State Corporation Commission of Virginia, and almost nobody noticed.

It was a charter for a new nonprofit. The language was the careful, weightless language of incorporation papers everywhere: a board structure, an operating mandate, an authorization to receive funds and to invest them. It authorized the new corporation to take money from the United States government, place that money into private technology companies, and return whatever it earned not to the Treasury but to a successor fund that would invest again. It was the kind of paperwork that announces nothing.

The charter did not use the word surveillance. It used the word innovation.

The distance between those two words, held open and load-bearing for the next twenty-five years, became one of the most consequential ambiguities in the history of American technology. Not because anyone was lied to. Almost every move the corporation ever made is, in some traceable form, a matter of public record. The ambiguity did something subtler than concealment. It changed the grammar. It took a government decision about what to buy and dressed it in the vocabulary of investment, portfolio, partnership, and exit, and in that vocabulary a strange thing happens to responsibility. A procurement has an author. A market does not.

The corporation was called In-Q-Tel. Its first and principal customer was the Central Intelligence Agency.

The word that did the work

Begin with the language, because the language is the mechanism.

When a government agency buys a thing, the act is legible. There is a line item, an appropriation, a committee that can ask why. When an agency invests in a thing, through an arm's-length nonprofit that speaks the dialect of venture capital, the same act becomes something the oversight machinery was never built to see. It is not classified, exactly. It is not commercial, exactly. It sits in a third category, and the third category is the whole invention.

This is the part most accounts of the surveillance state miss, because they go looking for a secret and the secret is not where the power is. The power is in the noun. Call a capability a weapons program and it triggers a century of law, hearings, and treaty. Call the same capability an innovation and it triggers a press release. The decision about which word applies is made early, quietly, by people who understand exactly what the choice does, and once the word is chosen the rest of the institutional world arranges itself around it. The philosopher Michel Foucault spent his career on this single observation: that the deepest form of power is not the power to decide an argument but the power to set the terms in which the argument can be had, to decide what counts as knowledge, what counts as a threat, what counts as normal. In-Q-Tel is that observation rendered as a balance sheet. It did not win the debate about intelligence-funded technology. It made sure the debate was filed under the wrong heading.

A procurement has an author. A market does not. That substitution, performed once at the level of vocabulary, is what let an intelligence agency become a founding investor in the instruments of everyday civilian life without ever appearing to be one.

The man who taught the agency to speak Silicon Valley

By the late 1990s the Central Intelligence Agency had a problem that money alone could not fix. Its in-house technology was running years behind the commercial world. The ability to search vast quantities of open text, to recognize a face or a pattern across a flood of images, to read a dozen languages at machine speed, was migrating out of government laboratories and into start-ups the agency had no way to reach. You cannot recruit a twenty-six-year-old machine-learning engineer with a federal pay grade and a polygraph. The talent was in a different world, and that world did not return the agency's calls.

The answer, arrived at under CIA Director George Tenet and pushed by an outside review of the agency's technology gap, was to stop trying to summon Silicon Valley and instead to become a creature it already recognized: an investor. The new fund would be chaired by Norman Augustine, the former chief executive of Lockheed Martin, a man who knew exactly how the defense world worked. And to run it day to day, they chose someone from the other side entirely.

Gilman Louie was a video-game designer. He had built flight simulators, including the Falcon series that a generation of teenagers flew on home computers. He had no intelligence background, and that absence was the qualification. The fund needed to look like Silicon Valley, talk like Silicon Valley, and sit across the table from founders as one of them rather than as a man from Langley in a bad suit. Louie was the translation layer made flesh. He could walk into a Mountain View conference room and the room would relax, because the person across the table spoke the language of term sheets and demo days, not of clearances and compartments.

The fund's legal form completed the disguise, and it is worth dwelling on because it is the structural keystone of the whole thing. In-Q-Tel was organized as a 501(c)(3) nonprofit. A nonprofit whose revenue arrives from a classified federal budget, and whose investments are disclosed when convenient and withheld when not, occupies a slot that the existing instruments of oversight simply do not fit. Congressional appropriators see a grant. Securities regulators see a private investor. Journalists see a charity. Each looks at one face of the thing and sees something ordinary, and the ordinary faces add up to something for which there is no name and therefore no committee.

The oversight gap is not a hole in a fence. It is a property of the architecture.

What the money actually bought

Here is where the usual telling goes wrong. People hear "the CIA funds tech companies" and reach for the size of the checks, and the checks are not large. In-Q-Tel began with thirty million dollars and has never written the kind of cheque that builds a giant. If the story were about capital, it would not be a story at all, because the commercial venture market dwarfs anything an intelligence nonprofit can spend.

The money was never the lever. Three other things were.

The first was the requirement that came attached. In-Q-Tel did not invest the way a normal fund invests, hunting only for a return. It invested to obtain a capability, and its deals carried the obligation to deliver what the company built into the hands of the intelligence community. The point of the investment was not the profit. It was the product, routed back to the customer.

The second was timing. In-Q-Tel arrived early, at the seed and the first round, the moment when a company is deciding what it is for. A fund that shows up at that moment does not just buy a slice of a finished thing. It helps decide which thing gets finished. Capital applied at the origin shapes the genome.

The third, and the most powerful, was certification. To have taken In-Q-Tel's money was to wear a mark that the rest of the national-security market could read instantly. It was not a security clearance in any legal sense. It was something more useful: a signal that the intelligence community's own venture arm had inspected this company and found it fit for sensitive work. To a government procurement officer weighing two vendors, the mark removed hesitation. To the company, it opened doors that stayed shut to competitors who lacked it. The investment functioned as a credential, and the credential was worth far more than the cash.

What the money bought, in other words, was not ownership. It was direction.

Keyhole, and the order reversed

In 2003 a small company in Mountain View called Keyhole was running out of money. It had built something startling for its time: software, called EarthViewer, that let a person fly over a three-dimensional model of the entire planet from a desktop, zooming from orbit to a street corner in a continuous motion. It was beautiful, expensive to run, and nearly bankrupt.

A call came from Virginia. In-Q-Tel invested, alongside the National Geospatial-Intelligence Agency, and the timing was uncanny. Within weeks the United States invaded Iraq, and a version of EarthViewer, adapted with the agencies' help, was being used to move through the terrain of Baghdad. Intelligence analysts used it to watch installations. Television networks, granted access in exchange for an on-air credit, used it to fly their viewers over the war. The same tool served the operations room and the evening news in the same week.

In October 2004 Google acquired Keyhole. In June 2005 Google released its software to the world under a new name. You have used it. It is called Google Earth.

The word that public discourse reaches for, to describe this passage from spy tool to household app, is commercialization. The word implies a tidy, neutral order: someone invents a thing, the market adopts it, it spreads. But the Keyhole sequence runs in the opposite direction. The capability was seeded by the state for an intelligence purpose, hardened in a war, and only then released into civilian life through a private acquisition. The government did not adopt a technology that the market produced. It seeded one, classified it, tested it in combat, and let it out the civilian door.

Once you have seen that inversion, you start to see it everywhere, and you cannot unsee it. The thing arrives in your life wearing the friendly face of a consumer product, and its first user was an analyst looking for a target.

Palantir, and the value of a mark

In its first years, a young company named Palantir Technologies took an early investment from In-Q-Tel. The reported figure was around two million dollars, with some accounts putting it lower, a rounding error against what the company would later raise and be worth. By the conventional arithmetic of venture capital, it should have been forgotten.

It was the most important money Palantir ever took, and not for the reason the number suggests.

The two million dollars functioned exactly as a credential is supposed to function. It told every agency in the national-security market that the intelligence community's own venture arm had evaluated this company and judged it fit for the most sensitive work there is. In a market where trust is the scarce resource and a misplaced contract can end a career, that signal was worth more than any sum. Palantir spent its early years not chasing commercial customers but embedding itself inside government, and when it finally listed on the New York Stock Exchange in September 2020, at a valuation near sixteen billion dollars, more than half of its revenue still came from government clients.

The two million dollars did not build Palantir. It determined what Palantir was allowed to become.

That sentence is the whole mechanism in miniature. The state did not need to own the company, run the company, or even fund the company in any meaningful proportion. It needed only to stand at the doorway at the beginning and decide who was allowed through. Direction, not ownership. A hand on the tiller at the moment the course is set, and then twenty years of the ship sailing on as if it had chosen its own heading.

And here an objection should be rising, because it is the right one. Maybe none of this proves anything. Maybe In-Q-Tel was only a shrewd early reader of a wave that was coming anyway, and it was the wave that built the world we live in, not the fund. Hold that thought. It is strong enough that the rest of this essay has to earn its way past it, and it will have to be answered in full before the end. But notice, already, what the objection quietly concedes: that the wave had a direction, and that someone was standing at the headwaters, deciding which boats were allowed to launch.

The program that showed its face

To see why In-Q-Tel's method mattered so much, look at what happened to the version that refused to use it.

In 2002 the Pentagon launched a program with a name of almost comic honesty: Total Information Awareness. It was run out of a new Information Awareness Office under Admiral John Poindexter, who described it as a Manhattan Project for counter-terrorism. Its ambition was to fuse the data trails of whole populations, financial, medical, travel, communication, into a single searchable field. And it announced itself. The office adopted a logo of an all-seeing eye set atop a pyramid, casting its beam across the globe, beneath the Latin motto scientia est potentia, knowledge is power. It was, in effect, a government program that drew a careful picture of exactly what it meant to do and hung the picture on the wall.

The picture was its undoing. Once the press saw the eye and read the name, the outcry was immediate and crossed party lines, and in September 2003 the Senate voted to strip the program of its funding. Total Information Awareness was, officially, dead.

It was not dead. Reporting in the years that followed traced its core architecture as it migrated quietly into the National Security Agency under a new and unremarkable code name, its capabilities preserved and, by later accounts, its privacy safeguards dropped along the way. The capability did not die. Only the name died, and the name died precisely because it had been allowed to become visible.

Set the two programs side by side and the lesson is exact. Total Information Awareness and In-Q-Tel wanted versions of the same thing: a population made legible to the state. One announced itself with a name, a logo, and a slogan, and was killed within a year by the simple fact of being seen. The other filed itself under innovation, dressed as a venture fund, and is still investing a quarter of a century later. The difference between them was never capability and never intent. It was visibility. The state learned, in public and in real time, that the surest way to keep an all-seeing program alive is to make certain no one can draw its picture.

The instruments of a legible population

Step back from any single deal and look at the shape of the whole portfolio, because the shape is the argument.

By 2016, In-Q-Tel had backed more than three hundred and twenty companies, over a hundred of them never publicly disclosed. The portfolio is not random. It clusters, with a consistency that is itself a kind of evidence, around a single capability described in many technical dialects: the capability to take a human population and make it readable. Geospatial mapping, so that every place can be seen. Facial recognition and biometrics, so that every face can be matched. Natural-language processing, so that every document and message can be parsed. Social-media mining, so that every public utterance can be collected and sorted. Network analysis, so that every relationship can be drawn.

The political scientist James Scott gave this ambition its name. He called it legibility: the deep drive of every modern state to make the people it governs countable, mappable, searchable, to flatten the dense and private texture of a society into something that can be read from the center. Scott was writing about cadastral maps and census rolls and the standardization of names. In-Q-Tel is the same impulse, updated to silicon and pointed at the digital exhaust of ordinary life. The fund did not buy a population's secrets. It seeded, across the whole stack, the instruments that turn a population into something legible in the first place, and it did so years before most citizens understood that their daily traffic was being made into a readable surface at all.

This is what makes the phrase surveillance economy more than a slogan. The watching is not bolted onto the digital economy from outside. It is woven into the foundation, present at the seed round, funded at the origin, because the same instruments that make a market efficient at knowing its customers make a state efficient at knowing its citizens. They are not two technologies. They are one, and it was seeded once.

The scholar Shoshana Zuboff named the commercial half of this arrangement surveillance capitalism: an economic logic in which the raw material is human experience itself, captured as behavioral data and refined into predictions about what people will do next. Her account begins with the business model, with the advertising giants that learned to turn attention and behavior into a product. What the In-Q-Tel record adds underneath it is the other half of the same foundation. The instruments that made surveillance capitalism possible were not only pulled into being by commercial demand. Many were seeded, at the origin, by an institution whose interest in a legible population was not commercial at all. The market wanted to know its customers. The state wanted to know its citizens. They reached for the same tools, and at the beginning, often, the same money built them.

Geofeedia, and the architecture working as designed

In 2016 the consequences arrived in a form no one could call abstract.

A company called Geofeedia sold software that scraped public social-media posts and sorted them by location, letting a buyer watch, in real time, what was being said from any chosen place on a map. Its customers included police departments across the country. In Oakland, in Baltimore, in Chicago, departments used it to monitor the demonstrations that followed police killings of Black Americans. The company's own marketing materials labeled activist groups as overt threats and offered law enforcement the chance, in its words, to stay one step ahead of the rioters. Facebook, Instagram, and Twitter had all granted it access to their data streams. In October 2016 the American Civil Liberties Union exposed the arrangement, and the platforms cut Geofeedia off within days. Stripped of the data access its product depended on, the company laid off about half its staff within weeks and never recovered.

Months earlier, a leaked document had revealed Geofeedia's other backer. It was an In-Q-Tel portfolio company.

Now hold the whole chain in view, because the chain is the lesson. An intelligence venture fund seeds a social-media monitoring tool. The tool is sold to police. The police point it at citizens exercising the most protected right they have, the right to assemble and protest. And at no point in that sequence did anyone at In-Q-Tel pick up a phone and direct a police department in Baltimore to surveil a demonstration. No order was given. No plan was executed. Each actor in the chain followed its own incentive, each transaction was legal, each step was a reasonable local decision, and the aggregate was a surveillance capability seeded by the state and aimed, by others, at the state's own dissenters.

No one was coordinating. That is the point.

The economist Friedrich Hayek spent his life on the difference between an order that is designed and an order that merely emerges from countless uncoordinated choices, and he insisted that an emergent order can look, from the outside, exactly as if a hand had arranged it, when no hand did. The surveillance economy is that kind of order. Looking for the villain who planned it is not just futile; it actively hides the mechanism, because the mechanism does not need a villain. It needs only a seed placed early, an incentive pointed forward, and a vocabulary that keeps anyone from having to take responsibility for the sum.

What 2013 made briefly visible

For a decade the architecture worked the way good architecture works, which is to say invisibly. Then, for a few weeks in the summer of 2013, the lights came on.

The documents Edward Snowden carried out of the National Security Agency described, among much else, a collection program called PRISM, running since 2008, under which the agency reached into the servers of nine of the largest names in American technology: Microsoft, Yahoo, Google, Facebook, PalTalk, AOL, Skype, YouTube, and Apple. The surveillance economy and the surveillance state, which the public had been encouraged to picture as separate worlds, turned out to run through the same pipes. The commercial platforms that gathered behavioral data to sell advertising were, through various legal instruments, also a collection surface for the state.

It would overclaim to say In-Q-Tel built PRISM. It did not, and the path from a seed investment to a classified collection program runs through many hands and many years. But the revelation lit up the shape the seeding had produced. The capacity to gather, resolve, and search the digital lives of an entire population had been built into the commercial layer at the foundation, with state capital present at the origin of many of its instruments, and once that infrastructure existed, the question of whether the state would also draw on it answered itself. You do not need to direct a river after you have dug the channel. PRISM was the channel, made briefly visible, before the water closed back over it and the country moved on to the next thing.

What was never funded

There is a second way to read a portfolio, and it is the one almost no one performs. You can read it by its absences.

For every category In-Q-Tel funded heavily, ask what it did not fund. The portfolio is dense with tools that make populations legible to institutions. It is conspicuously thin in the opposite direction: tools that make institutions legible to populations, or that protect a citizen's data from collection rather than enabling it, or that decentralize and encrypt and obscure rather than gather and resolve and reveal. The fund poured itself into the machinery of seeing and left the machinery of not-being-seen to others.

Discipline matters here, because absence is the most easily abused kind of evidence. The shape of a portfolio is not proof of intent; one can always construct an innocent story for any single gap, and a fund built to serve intelligence collection would tilt toward collection tools for reasons that require no dark motive at all. So treat this as what it is: a marked observation, a pattern in the documented record that is consistent with the thesis and worth naming, not a smoking gun. The investments are the hard evidence. The absences are the direction they point.

But name it we should, because the asymmetry is the quiet center of the whole story. A society's tools are not neutral. Someone funds the ones that get built early and well, and the ones that get built early and well become the infrastructure everyone else has to live inside. When the seed capital for a generation of foundational tools comes overwhelmingly from an institution whose mission is to see, the built world tilts, gently and permanently, toward being seen.

The fund that never empties

There is one more feature in that original four-page charter, and it is the one that turns a program into an institution.

In-Q-Tel was built to return its proceeds not to the Treasury but to itself. The gains from a successful exit flow back into the fund to be invested again. It is an evergreen machine, designed never to run dry, designed to outlast any single director, any single administration, any single decade's understanding of what it is for. Tenet left. Louie left. Presidents came and went, intelligence priorities rotated from terrorism to great-power competition to artificial intelligence, and through all of it the fund kept seeding the next generation of instruments, financed in part by the success of the last.

This is the feature the historian Fernand Braudel would have recognized at once. Beneath the fast surface of events, the directors and the scandals and the news cycles, sits a slow layer that barely seems to move and quietly sets the terms for everything above it. In-Q-Tel has become that slow layer for the civilian technology stack. It is the patient, permanent presence at the origin of things, the investor that was already in the room before the founders arrived and will be in the room after they are gone. In the end it is the fund that shapes the market, and the market that shapes that fund, each financing the other across a quarter of a century with no one having to decide, on any given morning, to keep it going.

And so the question that the charter raises, twenty-five years on, is no longer whether an intelligence agency funds technology. It plainly does, and that alone would be unremarkable. The question is harder and it should be marked as an open one, because it lives at the edge of what the documents can prove: at what point does a state's venture fund stop being a buyer of tools and become an author of the civilian world those tools are made of? When the seed capital for a generation of the instruments through which a society sees itself has flowed, early and quietly, from the same patient source, the line between procurement and architecture begins to blur. No one announces the crossing of it. It is the kind of line a system passes without noticing it has passed.

The strongest case against this reading

The objection planted earlier now deserves its strongest form, because there is a real one, and a piece like this is only worth as much as its ability to survive it.

It runs like this. The causation is backward. These technologies were going to be built regardless; the commercial incentives to map the earth, recognize faces, and mine social media were enormous and entirely independent of any intelligence agency. In-Q-Tel did not create the surveillance economy; it merely noticed, early and shrewdly, a wave that was already coming and bought a surfboard. The sums are trivial against the trillions the private market poured into the same technologies. Google Earth would exist if Keyhole had taken any other investor's money; Palantir's founders were already building their company; the social-media giants needed no encouragement to harvest data, because harvesting data was their business model. On this reading In-Q-Tel is a footnote, a clever government program that rode a commercial tide it did not cause, and to make it the author of the surveillance economy is to mistake an early passenger for the driver.

Much of that is true, and the true part must be conceded fully. The commercial appetite was real and would have built most of this regardless. The sums were small. The wave was coming.

And yet the argument answers a claim that was never made. The thesis was never that In-Q-Tel's money created these technologies. It is that the state's seed capital, applied at the origin, with the requirement to share, the timing to shape, and above all the certification to anoint, helped determine which capabilities matured first, which companies became trusted infrastructure, and which direction a neutral-seeming toolset was pointed. A wave needs no help to arrive. But the question of which surfers reach the shore, and which beach they land on, is decided by exactly the kind of early, structural advantage that a credential and a steering hand confer. Scope the claim honestly and it holds: not that the intelligence state invented the surveillance economy, but that it was present at the creation, shaping the genome, in a way that the language of innocent commercialization is specifically designed to make invisible.

The word on the charter

Return, at the end, to the four pages filed in Virginia, and to the choice buried in them.

The men who built In-Q-Tel were not cartoon conspirators. They were solving a genuine problem, a real and defensible gap between what the intelligence community could do and what the world now demanded of it, and they solved it with a piece of institutional design so elegant that we are still living inside it. The elegance was the danger. They did not hide the program. They did something far more durable. They named it correctly, in the only sense that matters: they chose, at the very first step, the word that would keep the program from ever being seen for what it was.

They could have called it a procurement vehicle for intelligence technology. It was one. Instead they called it innovation, and innovation has no enemies, no oversight committee, no treaty, no author to hold to account. The whole twenty-five-year architecture, the inverted order of Keyhole, the credential that made Palantir, the instruments of a legible population, the seed that found its way to a police department in Baltimore, all of it descends from that single act of naming.

Gilman Louie, the game designer who had taught the agency to speak Silicon Valley, left the fund in the middle of the decade and went back to ordinary venture capital, his translation work done. The fund did not need him anymore. It had never needed any particular person; it needed only the structure, and the structure he had helped install kept running without him, exactly as it was built to.

We were told a story about the market discovering wonderful new tools and generously sharing them with us. The truer story runs the other way, and it was always available, sitting in plain public language that no one was meant to read closely. The most powerful surveillance system ever built was not imposed on a free market from outside. It was seeded inside it, at the beginning, by an investor who was careful, above all, about which word went on the charter.

Evidence Map

Facts, interpretations, forecasts, and disconfirming signals.

Core claim. The surveillance economy was not merely adopted by the state after the market built it; it was seeded by the state at the origin, through In-Q-Tel, the CIA's venture arm, whose real leverage was never its small capital but its power to certify companies, require that capabilities be shared, and steer which instruments of population-legibility matured first, all under a vocabulary ("innovation," "investment") that made the procurement look like ordinary market behavior and erased its author.

Evidence level. Facts (high): In-Q-Tel founded 1999 under CIA Director George Tenet with about 30 million dollars, chaired by Norman Augustine, first CEO Gilman Louie; 501(c)(3) structure; Keyhole/EarthViewer In-Q-Tel and NGA investment in 2003, used during the 2003 Iraq invasion, acquired by Google in 2004, released as Google Earth in 2005; In-Q-Tel's early investment in Palantir (reported around 2 million dollars), which listed in 2020 with more than half its revenue from government; more than 320 portfolio companies by 2016, over 100 undisclosed; Geofeedia an In-Q-Tel company, exposed by the ACLU in 2016 for police monitoring of Black Lives Matter protests, cut off by the platforms and collapsed soon after. Interpretation (medium, marked): the reading of certification and timing (not capital) as the determining variable; the "vocabulary erases authorship" thesis; the legibility-portfolio pattern; and the open, marked question of whether a state venture fund at this scale and permanence has crossed from procurement into authorship of the civilian technology stack. Contested/soft (marked): the exact Palantir figure (reported 1.25 to 2 million) and the weight of the portfolio's absences, which are signal, not proof.

What would confirm this. Further declassified or leaked records showing In-Q-Tel's share-back requirements and steering at the seed stage; documented cases where the In-Q-Tel credential decided a procurement; continued portfolio concentration in population-legibility tools relative to citizen-protection tools.

What would disprove this. Evidence that In-Q-Tel's investments imposed no capability-sharing obligations and conferred no procurement advantage; a portfolio that is in fact balanced between collection tools and privacy/decentralization tools; cases where In-Q-Tel-backed companies systematically failed to reach government work, showing the credential carried no weight.

Watchlist. In-Q-Tel's current investments in artificial intelligence and biometrics; new disclosures of undisclosed portfolio companies; whether the evergreen reinvestment structure continues to outlast administrations; the migration of intelligence-seeded tools into consumer products.

Jerry van der Laan writes The Manifest Archive, daily forensic essays on power, language, and the systems that shape what we are allowed to see as reality. He traces the structures beneath them.