Portraits — American and Comparative

Everything argued so far has been argued in the abstract, and deliberately so. The trap described in this book is structural. It does not depend on the character of any particular person, and it would persist if every individual currently occupying its commanding heights were replaced tomorrow.

But abstractions can be tested. If the preceding chapters are right, we should be able to look at the world as it exists and find the pattern running: private infrastructure making decisions that once belonged to states, observation and narrative accumulating under single owners, governmental judgment migrating into privately held software. What follows are three American portraits drawn from the public record as of 2026, followed by two comparative non-U.S. cases that demonstrate the same structural logic operating under different political and ownership regimes.

A caution before beginning. None of these portraits is an accusation, and each ends by conceding its subject's best argument, because the subjects have serious arguments. That is the point. If the danger required villains, it would be easy to spot and easy to stop. The danger described here is compatible with every one of these actors being exactly what they believe themselves to be.

The Man Who Owned the Battlefield's Signal

In the fall of 2022, three United States senators wrote to the Pentagon with a question that would have been incomprehensible a generation earlier: had a private citizen unilaterally shaped the outcome of a military operation in an active war between nation-states?

The citizen was Elon Musk. The operation was a Ukrainian naval drone attack on Russia's Black Sea Fleet near Crimea, which, according to Musk's biographer Walter Isaacson, failed to proceed as planned because Musk declined to enable Starlink satellite coverage in the region.

Musk's account differs in emphasis. He has said he denied a request to extend coverage rather than disabling an active mission, and that he acted out of genuine fear of nuclear escalation. But the dispute over details obscures the fact that both versions describe the same underlying reality. A decision that would ordinarily belong to a president, a war cabinet, or at minimum a defense department was made instead by one man, consulting his own judgment, accountable to no electorate. As Senator Jack Reed put it at the time, neither Musk nor any private citizen can have the last word on U.S. national security. Yet at the moment of the incident, Starlink had no Pentagon contract covering Ukraine, and therefore no directive from U.S. officials. The last word was, in fact, his.

This is the power curve made flesh. Starlink was not built as a weapon. It was built as a consumer internet business. But when Russia destroyed Ukraine's communications infrastructure in 2022, SpaceX's satellite network became, almost overnight, the connective tissue of an entire war effort: battlefield awareness, artillery coordination, drone operations. A space policy analyst at the Center for Strategic and International Studies observed that activating Starlink in Ukraine was a policy decision the U.S. government never made. The government could not have made it. It did not own the infrastructure.

What followed traced the arc this book describes. In early 2025, Reuters reported that U.S. negotiators had raised Ukraine's Starlink access as leverage in negotiations over a critical minerals agreement, a claim Musk denied, but one whose mere plausibility rattled Kyiv, because everyone understood that the leverage was real whether or not it was being exercised. By 2026, reporting indicated Musk was still declining to permit Starlink use for Ukrainian strikes inside Russian territory, despite direct appeals from President Zelensky, who reportedly asked the American president to intervene with Musk personally. Read that sentence again: the head of one state asking the head of another state to lobby a businessman.

Meanwhile, the businessman's leverage compounded. In February 2026, SpaceX absorbed Musk's artificial intelligence company xAI in a transaction that valued the combined entity above a trillion dollars, folding a rocket company, a global satellite network, an AI laboratory, and a social media platform into a single corporate structure. The subsequent public offering, the largest in history, briefly made Musk the world's first trillionaire on paper. The S-1 filing disclosed that through supervoting shares he would retain roughly eighty-five percent of voting power, with the filing stating plainly that he would control the outcome of essentially every matter requiring shareholder approval. Public investors were invited to buy economic exposure. Control was not for sale.

None of this required villainy, and this portrait should not be read as an indictment of Musk's character. His stated reasoning in the Crimea episode, fear of escalation toward nuclear war, is a consideration any responsible official would weigh. He may even have been right. That is precisely the trap. The question is not whether Elon Musk made a defensible call. The question is why the call was his to make, and what happens when the next call, made by him or by someone like him, is wrong. China grasped the implications early: Beijing pressured Musk for assurances about Starlink's availability over Taiwan, and accelerated its own state-controlled satellite constellation in response. Foreign governments now conduct diplomacy with a man, because the man owns what states need.

A single individual now sits atop the world's dominant orbital launch capability, a communications network that militaries depend on, a frontier AI lab, and a major information platform, and holds, by design, voting control over all of it. Whether he uses that position wisely is a question about him. That the position exists is a question about us.

The Cheerful Panopticon

In September 2024, at a meeting with financial analysts, Oracle co-founder Larry Ellison described the future he expected artificial intelligence to deliver. Cameras everywhere: dashboards, doorbells, drones, police uniforms, with AI processing the footage continuously.

"Citizens will be on their best behavior," he predicted, "because we are constantly recording and reporting everything that's going on."

What makes the remark remarkable is not its dystopian content. Science fiction has imagined such worlds for a century. What makes it remarkable is that it was not a warning. It was a product roadmap, delivered approvingly, by a man with the means to build it.

Consider what Ellison assembled in the two years that followed. Oracle's cloud business became foundational infrastructure for the AI boom, hosting workloads for U.S. intelligence agencies and holding health-technology contracts touching the medical records of tens of millions of Americans. In 2025, Skydance Media, funded by Ellison and led by his son David, completed its merger with Paramount Global, giving the family effective control of CBS, CBS News, Paramount+, and a portfolio of cable networks; reporting has consistently suggested the elder Ellison, holding roughly half the merged company, is the true center of gravity. The family then pursued Warner Bros. Discovery, which would add HBO and CNN to the portfolio. And in the government-brokered restructuring of TikTok's American operations, the president announced that Ellison and Oracle would play a central role, placing the recommendation algorithm that shapes what roughly 170 million Americans watch under the influence of the same man who mused about a society kept well-behaved through constant observation.

Each acquisition, viewed alone, is an ordinary business story. An aging founder diversifies into media; a cloud company wins government contracts; a strategic asset finds politically acceptable ownership. That is exactly how the previous chapter on surveillance predicts it will feel. Surveillance infrastructure is not imposed. It accretes, through deals, conveniences, and transactions that each seem unremarkable until the accumulated whole becomes something no society ever voted for.

The Ellison case adds a dimension the Flock Safety example in Chapter Five cannot: vertical integration of observation and narrative. A camera network watches behavior. A media empire shapes the interpretation of behavior. A social platform's algorithm decides which interpretations spread. A cloud business holds the data underneath all of it. There is no evidence these holdings are being coordinated toward the panopticon Ellison described, and fairness requires saying so plainly. But this book's argument does not depend on coordination. It depends on capability. The relevant fact is that one family now could shape, to a degree without obvious precedent, both what is seen and how it is understood, and that the only safeguard against misuse of that position is the family's own judgment.

Ellison, in his ninth decade, has been rewarded across a sixty-year career for trusting his instincts about where technology is heading. He has usually been right. The next chapter explains why that is precisely the problem.

The Company That Became the State's Nervous System

If Musk illustrates infrastructure and Ellison illustrates observation, Peter Thiel's Palantir Technologies illustrates the quietest and perhaps most consequential pattern of all: the migration of governmental judgment itself into privately owned software.

Palantir was founded in 2003 with early backing from the CIA's venture arm, and for its first two decades it was a niche, often unprofitable contractor. Then the curve bent. Federal contract revenue grew from roughly four million dollars in 2009 to over half a billion dollars in 2024, and then nearly doubled again in 2025. In mid-2025, the Army consolidated dozens of separate agreements into a contract vehicle worth up to ten billion dollars over a decade. In early 2026, the Department of Homeland Security signed a roughly billion-dollar agreement for software supporting immigration enforcement. Palantir's systems now run inside the Departments of Defense, Homeland Security, Treasury, State, Health and Human Services, and a dozen other agencies. The company that once struggled to win business has become something close to the federal government's operating system.

Recall the scenario from Chapter Six: a commander ratifying a chain of machine-generated conclusions, the human loop preserved in form and emptied of substance. Palantir's ImmigrationOS contract with ICE offers the domestic civilian version. The system is designed to integrate data across sources, track individuals, and help prioritize enforcement targets. Every arrest still involves human officers making human decisions. But which names surface, in which order, with which risk flags attached: that is increasingly the software's contribution. Members of Congress have demanded answers about whether the resulting apparatus constitutes what two representatives called a mass surveillance ecosystem. Thirteen former employees published a letter condemning the company's work for the administration. The company's CEO, for his part, has been forthright rather than evasive: asked whether adversaries are surveilled using data flowing through Palantir's products, Alex Karp answered one hundred percent, and said he completely supported it.

The incentive machinery around the company completes the picture. Palantir sharply increased its federal lobbying as contracts expanded. It donated to the president's inaugural committee. A senior White House official who served as chief architect of the very immigration policies Palantir's software operationalizes was found to hold a personal stake in the company's stock, later sold, with the administration dismissing concerns, but a tidy illustration of the previous chapter's thesis that no conspiracy is required when interests simply align. In August 2026, in a development a satirist would have rejected as too on-the-nose, the State Department enlisted Palantir, a surveillance firm, to advise the government on digital freedom and countering surveillance.

And here, too, the steelman deserves its due, because Thiel and Karp have one. Their argument, made openly for years, is that Western governments face genuine adversaries, that software superiority is now inseparable from military superiority, and that Silicon Valley's squeamishness about defense work amounts to free-riding on the security it privately disdains.

This is a serious position, and events since 2022 have strengthened rather than weakened it.

This book's response is not to dismiss the argument but to complete it: even granting that democracies need companies like Palantir, a democracy that cannot audit, constrain, or replace its own nervous system has not procured a service. It has acquired a dependency. When a single firm's platforms mediate how the state sees, who gets flagged, who gets found, who gets targeted, the question "is the software aligned?" becomes inseparable from the question "is the state still the principal, or has it become the client?"

Thiel himself has written that he no longer believes freedom and democracy are compatible. One need not treat the line as a confession of intent to notice the asymmetry it reveals: the citizens whose data flows through his company's systems were never asked whether they agreed.

Comparative Case: The Surveillance Stack as National Infrastructure

The American cases above show private capability becoming public infrastructure under conditions of formal legal separation and residual democratic contestation. Outside the United States the same technical and economic forces operate under different ownership and political arrangements, often with even thinner external constraints.

China’s video-surveillance and AI-analytics industry provides the clearest large-scale illustration. Firms such as Hikvision and Dahua grew into global market leaders in networked cameras and related hardware, while companies including SenseTime, Megvii, and iFlytek specialized in the computer-vision and speech models that turn raw video and audio into searchable, actionable data. These companies have been deeply integrated into domestic “safe city,” “city brain,” and public-security platforms. Procurement documents and public reporting show AI-enabled systems deployed for real-time behavioral analysis, multi-camera tracking, and predictive alerting at municipal and provincial scale. The same firms have exported hardware, software, and turnkey platforms to governments across the Global South and beyond. The resulting systems are not fringe experiments; they constitute a significant fraction of the world’s installed surveillance capacity.

What distinguishes the Chinese pattern is the relative absence of institutional distance between the commercial developers and the state agencies that deploy their products. State ownership stakes, military-civil fusion policies, and explicit national-security guidance reduce the autonomy that private firms in liberal democracies can claim. The incentive structure is correspondingly clearer: the largest buyer is the state, the largest use-cases are security and social management, and the firms that best serve those priorities receive capital, talent, and market access. The same competitive logic described in Chapter Seven operates, only with the state as both customer and, in important respects, co-owner.

The steelman is straightforward. Chinese officials and company executives argue that advanced surveillance improves public safety, deters crime, and enables efficient governance in a vast and complex society; that Western criticism is selective and often geopolitical; and that every major power is racing to master the same dual-use technologies. The argument has force. Yet it does not erase the structural fact: capability of extraordinary reach has been concentrated in organizations whose accountability runs primarily upward to political authority rather than outward to the populations under continuous observation. Once the cameras and models are installed, the institutional questions that Chapter Five treats as open—who may query, for how long, under what audit—are answered by the regime that owns the stack. Export of the technology often carries elements of that governance model with it.

Comparative Case: Sovereign Capital and the AI Infrastructure Race

A second non-U.S. pattern appears in the Gulf, where sovereign wealth funds controlled by ruling families have become decisive capital providers for AI infrastructure and frontier model development. The United Arab Emirates’ G42, chaired by a senior member of the ruling family who also serves as national security adviser, has assembled an AI operating company spanning cloud, models, health data, and large-scale compute. Partnerships with Microsoft, OpenAI, Nvidia and others, combined with multi-gigawatt data-center plans (including the Stargate UAE campus), position the firm as a major non-U.S. node in the global AI stack. Saudi Arabia’s Public Investment Fund, through vehicles such as HUMAIN, has likewise committed tens of billions to domestic compute capacity, chip partnerships, and AI applications tied to national diversification goals.

In these cases the concentration of capital is not the product of successive entrepreneurial successes by private individuals; it is the deliberate allocation of state-controlled hydrocarbon wealth into strategic technology. Accountability flows to the sovereign, not to public shareholders or an electorate in the liberal-democratic sense. The same structural features reappear: enormous deployable resources, rapid decision cycles unconstrained by multi-year legislative process, and the capacity to shape who builds what and who gains access. The steelman is again serious. Gulf states argue that they are converting finite resource wealth into future-proof technological capacity, reducing dependence on oil, and positioning themselves as neutral or aligned partners in a multipolar AI landscape. The claim has merit. Yet the asymmetry remains: decisions of planetary technological consequence are being taken by a very small number of people whose power derives from control of sovereign capital rather than from broad democratic authorization, and whose error-correction mechanisms are correspondingly narrow.

What the Portraits Share

Three American men of very different temperaments, and two non-U.S. configurations of capital and state power. What they share is not ideology or nationality. It is position: each sits at a junction where concentrated capability has quietly become public or quasi-public infrastructure, and each answers, in the end, to a very small set of principals.

That is the condition the next chapter names.