Yesterday I wrote about the auction — the real-time bidding machinery that describes you, hundreds of times a day, to whoever is willing to pay attention. I want to stay with that subject one more day, because of a news item that keeps circling it.
The news item
In June, Palantir and Zeta Global announced a partnership. If you follow defense reporting, you know Palantir: it builds the data-integration and analysis software used by intelligence agencies and military organizations on several continents. If you follow marketing, you know Zeta: it runs one of the industry’s large consumer identity graphs, assembled from signals on roughly 245 million people in the United States. The deal puts Zeta’s data cloud on Palantir’s Foundry platform and aims the combination at enterprise marketing. That is the whole story, and the trade press covered it mostly as a business story, which is fair. I am not claiming anything dark about the companies or the deal.
What I notice is what the combination is for: knowing an individual well enough to predict what they will do next, and acting on that prediction automatically. Marketing calls that personalization. The intelligence community has spent decades buying versions of the same capability, and the fact that one of its main suppliers is now selling an architecture for marketing should tell you something about how similar the two problems turned out to be.
Hesitation first
I want to acknowledge something before going further. I hesitate to write about this at all, because where the words “surveillance” and “advertising” appear in the same sentence, readers reasonably brace for a conspiracy theory. I have spent years working with data at scale, and that experience is exactly why I keep circling back to this subject anyway. It is not that anything secret is happening. It is that the ordinary, documented, entirely public machinery is powerful enough that you do not need a conspiracy for the outcomes to be strange.
This is not a plot twist
Intelligence agencies have been paying attention to the advertising ecosystem for years — as customers, not as saboteurs. It is public record that data brokers have sold location and behavior data drawn from the ad ecosystem to government agencies, and that agencies have used those purchases to get at information the law would not let them collect directly with a warrant. Researchers have estimated that about a thousand dollars of targeting data can be enough to de-anonymize an individual. The advertising network is, among other things, the largest commercially available surveillance dataset on earth, and it is open to anyone who registers as a buyer. It always has been, and it would be strange if the professionals whose job is knowing things had ignored it.
This year also produced the Discover-Tech reporting I covered yesterday: an ad-tech company whose ordinary-seeming bidding activity doubled as a targeting pipeline for mercenary spyware, with ads rendered on networks belonging to the Pentagon, NATO, and the International Atomic Energy Agency. That case and the Palantir–Zeta deal sit at two ends of the same fact. The ad ecosystem knows individuals so well — so precisely, so quickly, across so many apps and sites — that everyone with a professional interest in knowing people ends up in the same auctions, or in business with the companies that run them.
Both sides of the same desk
And none of this requires anyone to be a villain. The machinery was built to sell things. The thing that makes it easy for you to buy the album you have been half-remembering for years, or to finally see the car you actually want instead of a generic sedan ad, is real. The same machinery, pointed by a different buyer, is how someone learns where you live, where you worship, and what you are afraid of. It is the same bid request. The difference is only who is asking and why.
The book is largely about this symmetry. Chapter 4, “They Know You Inside and Out,” walks through the machinery of recognition — how the ad ecosystem connects your visits into a single person across sites and apps, and what the resulting profile assumes about you. Chapter 3 covers the part people underestimate: the system is not just watching you, it is testing you, adjusting what it shows you and measuring what that does to you, in real time. And chapter 8 is about what you can actually do — take charge of what these systems remember about you, instead of pretending you were never in the auction to begin with.
The practical point
I am not writing this to scare anyone, and I am not claiming to have uncovered anything. Most of what I have described is in regulatory filings and university lab reports. My point is smaller and more useful: the same infrastructure that recommends your next album is infrastructure that other people can query about you, and it is worth being aware of that as a fact of modern life rather than as a plot. You do not have to go off the grid. It is enough to know that the convenience and the exposure are the same machine, and to make your choices about it deliberately.
If any of this is new to you, the book is the longer, calmer version of the same argument. It is available in paperback and Kindle.