cryptogon.com

Depending where you fall on the spectrum between civil liberties absolutism and homeland security lockdown, Palantir’s technology is either creepy or heroic. Judging by the company’s growth, opinion in Washington and elsewhere has veered toward the latter. Palantir has built a customer list that includes the U.S. Defense Dept., CIA, FBI, Army, Marines, Air Force, the police departments of New York and Los Angeles, and a growing number of financial institutions trying to detect bank fraud. These deals have turned the company into one of the quietest success stories in Silicon Valley—it’s on track to hit $250 million in sales this year—and a candidate for an initial public offering. Palantir has been used to find suspects in a case involving the murder of a U.S. Immigration and Customs Enforcement special agent, and to uncover bombing networks in Syria, Afghanistan, and Pakistan. “It’s like plugging into the Matrix,” says a Special Forces member stationed in Afghanistan who requested anonymity out of security concerns.

…

Palantir’s engineers fill the former headquarters of Facebook along University Avenue in the heart of Palo Alto’s main commercial district. Over the past few years, Palantir has expanded to four other nearby buildings as well. Its security people—who wear black gloves and Secret Service-style earpieces—often pop out of the office to grab their lunch, making downtown Palo Alto feel, at times, a bit like Langley.

…

The origins of Palantir go back to PayPal, the online payments pioneer founded in 1998. A hit with consumers and businesses, PayPal also attracted criminals who used the service for money laundering and fraud. By 2000, PayPal looked like “it was just going to go out of business” because of the cost of keeping up with the bad guys, says Peter Thiel, a PayPal co-founder.

The antifraud tools of the time could not keep up with the crooks. PayPal’s engineers would train computers to look out for suspicious transfers—a number of large transactions between U.S. and Russian accounts, for example—and then have human analysts review each flagged deal. But each time PayPal cottoned to a new ploy, the criminals changed tactics. The computers would miss these shifts, and the humans were overwhelmed by the explosion of transactions the company handled.

PayPal’s computer scientists set to work building a software system that would treat each transaction as part of a pattern rather than just an entry in a database. They devised ways to get information about a person’s computer, the other people he did business with, and how all this fit into the history of transactions. These techniques let human analysts see networks of suspicious accounts and pick up on patterns missed by the computers. PayPal could start freezing dodgy payments before they were processed. “It saved hundreds of millions of dollars,” says Bob McGrew, a former PayPal engineer and the current director of engineering at Palantir.