We are flowing down a hallway in an office building in Tysons Corner, a little cloud of government functionaries and a single journalist. We make a turn to the right, pass through a serious security door, and then come to a stop. Like bears in the forest, the government officials mark their presence by carefully rubbing their magnetic ID tags on a featureless white box mounted on the wall. Somewhere an invisible electronic brain cogitates, and permission is granted. We enter the Operations Center of the Financial Crimes Enforcement Network, or FinCEN.
The Op Center is a midsize L-shaped room with a bank of desktop computers and related equipment against the long wall, everything whirring softly. The most ominous detail is the most humble: a row of plain, government-issue wall clocks set in Strangelovian fashion to the world’s various time zones. This is the kind of room Hollywood set designers choose for fighting off alien invasions or managing secret global governments.
The staff belies the setting. The half-dozen computer operators in the room resemble either grad students (track suits and long hair) or government wonks (bad suits and short hair). The sidearm of choice is a plastic pocket protector.
I am set down with Joanne Brown, a cheery, brown-haired young woman who drives a computer for a living. Brown is an analyst for FinCEN, the youngest and most obscure law enforcement agency in the federal government. Barely 6 years old, FinCEN’s job is to detect criminal money laundering by cash-flush drug dealers, medical-fraud artists, corporate embezzlers, and anyone else who puts the profits of sin into a bank account.
FinCEN’s central mission is simply to follow the money. En route to this original goal, the agency has acquired some additional functions that are a bit more troublesome to people who keep an eye on privacy issues. With access to millions of secret financial records, immeasurable gigabytes of public information on everything from driver’s licenses to home liens, and an entree into the highly restricted records of the FBI and other acronymic federal police shops, FinCEN has become something of a full-service information clearinghouse for the government. It’s a major convenience for people with badges, a one-stop shopping destination for local, state, and federal law enforcers eager to learn everything they can about suspected criminals.
With those kinds of hellacious resources, FinCEN has broadened the typical menu of enforcement beyond busting drug dealers and bunk artists; the money-sniffing agency has snared doctors who have bilked Medicare, tracked down the U.S. assets of Iraqi officials, and traced payments made to suspects in the World Trade Center bombing.
In the main, there’s nothing new about FinCEN; government keeping information on people is an act as old as government itself. But FinCEN has given the process a high-tech kick with the application of artificial intelligence (AI), electronic monitoring of high finance, and big-time number crunching.
Naturally, some folks—from civil rights advocates to nutball minutemen huddled over their latest militia newsletter—find all this a little scary. Ever since its inception, FinCEN has been accused of snooping into our lives too deeply and of posing a threat to the foundation of privacy in American life. I came to FinCEN and the Op Center to see which vision rings true—whether the agency is an ominous Big Brother or just a shiny new marvel of law and order.
I wiggle in my chair nervously while Brown settles into her keyboard. Naturally, the subject of our first data hunt will be someone close at hand.
“If you would be kind enough to volunteer some information about yourself,” Brown says amiably, “I’ll show you what we have on you.”
The point of almost all crime is to make money. But what is money, exactly? All serious money loses its paper form at one point or another in its life and becomes a pixel, a dot of light traveling down a wire or riding a microwave across the landscape or even bouncing off a satellite. This abstraction of money is a paradigm shift as profound as the centuries-long switch from precious metals to paper, yet it is occurring far more quickly. Last year’s fantasies, like digital cash and encrypted Internet commerce, are now scheduled into someone’s fourth-quarter product line. Financial liquidity is approaching the speed of light.
FinCEN tracks the blips. In a wired world, federal law enforcement is trying to catch up with criminals who shift their ill-gotten gains from account to account—or country to country—in seconds. There is a vast, international trade in laundering money. Just how vast is unknown, but the Treasury Department’s best guestimate is that at least $300 billion dollars is laundered worldwide every year, an amount that continues to grow with the unlimited genius of human greed.
Laundering money simply means taking the profit of crime—whether from cocaine sales, bogus insurance claims, high-tech embezzling, or low-tech bank robbing—and hiding its origins. The vial of crack goes out one hand, a crumpled $20 bill comes into the other hand. To a Cali Mafioso, a Mexican wholesaler, or an American middleman, great bundles of dirty $20 bills are a liability, not an asset. The twenty must be converted, somehow, into a blip in an account that can be shifted around, used to buy real estate or airplanes or an investment portfolio.
The ways of hiding money are innumerable, while the ways of finding it multiply with every new line of program code at FinCEN. The agency is headquartered in the “doughnut building,” so named because of the six-story concrete O-ring incorporated into its façade. Except for the doughnut, the building is another anonymous glass tower at the intersection of asphalt and information highways in Tysons Corner. Like the interchangeable service-sector offices surrounding it, the doughnut building is a paradigm of the digital age, an Edge City complex where the doors are opened with magnetic keys, fabric cubicles can be rearranged overnight, and the twin pillars of the local infrastructure are the electrical plug and the phone jack. FinCEN blends perfectly into this ephemeral world of plastic-card commerce and liquid money.
FinCEN has three main functions: First, it regulates data reporting in the banking system; second, it has the “support” function of responding to requests from law enforcement agencies for profiles of suspect individuals and businesses; and last, FinCEN has been given the operational mission of “proactive” investigation, a fancy way of saying that rather than waiting for the DEA or the Customs Service to phone up a request for data on an individual suspect, FinCEN is actively scanning enormous amounts of data for suspicious financial patterns.
It’s a big job. About one-third of the global money-laundering trade, or perhaps $100 billion dollars a year, is “washed” through the American financial system. The dirty money may be buried in the profits of a legitimate business, deposited in a relative’s bank account, shifted into and out of the country overnight, or tucked behind a shell of paper corporations. More than a trillion dollars is wired through New York City every day, and there are literally thousands of banks in America, plus more than 200,000 other “nonbank” financial institutions as diverse as casinos and foreign-exchange shops. All of them could be used by money launderers. FinCEN’s job is to pluck a signal from all this white noise, a thread from the broad tapestry of legitimate commerce.
Although FinCEN has gained little notice from the public—many civil rights experts I called had never heard of it—the agency has quietly steered American law enforcement into uncharted territory. The tiny agency, with a staff of just 200 and a puny budget of about $22 million, has developed an ability unprecedented in government circles to compile dossiers on ordinary Americans, collating information from commercial databases, law enforcement rap sheets, corporate records, reports from intelligence agencies, and bank transactions into comprehensive profiles of suspects.
Is this Big Brother’s own headquarters? The few people who have heard of FinCEN—an odd coalition of right-wing neo-Luddites and left-wing technophiles—seem to think so. Wired magazine showed its usual restraint by headlining a feature on FinCEN, “Big Brother Wants to Look Into Your Bank Account.”
Let’s keep one foot on the ground. If you are looking for black helicopters on the roof of the doughnut building, you are missing the real story inside. There is no conspiracy to peer into your ATM transactions or snoop through e-mail; instead, the steady evolution and convergence of computer technologies has created two new challenges to our rights.
The first is the convergence of all transactions—credit card applications, magazine subscriptions, insurance claims, driver’s licenses—into all-encompassing profiles of every American. These public sources of information have long existed, but they have never been so comprehensive, nor so quickly and easily synthesized into dossiers. Thanks to the CD-ROM and a free-market economy, the same information is available to the Mafia, L.L. Bean, or the Girl Scouts, and arguably just as insulting to our privacy whoever has it. (In fact, as I discovered after reporting this article, even this paper has a database with some of that information on it. It won’t be long until everyone has a file on everyone else.)
The second and more profound challenge comes from FinCEN’s access to secret banking records. The agency’s task of proactive investigation sounds rather dry in practice, but it holds the nauseating potential to reverse a crucial tenet of our justice system by turning the presumption of innocence into a hunt for guilt. Police were once expected to have suspicion against an individual citizen before peering into his life. The technological abilities of FinCEN are slowly reversing that tenet of law enforcement: Now we investigate not because we should, but because we can. In the age of computer crime, we are, all of us, innocent and guilty, subject to examination by a computer program that leaves the judgment of suspicion to the silent calculations of an algorithm.
Mark Rotenberg, director of the Electronic Privacy Information Center (EPIC), is one of a small band of civil rights and computer privacy advocates who believe that this obscure government agency has gone too far, describing FinCEN as “off the rails.”
Rotenberg pauses after saying that. Given how much new territory FinCEN is exploring, and how quickly and powerfully it is moving forward, he decides that the image of derailment is not quite right.
“Perhaps I should say they are heading in some new direction,” he says, “laying their own tracks as they proceed.”
Back in the L-shaped Op Center, the cheery Joanne Brown is waiting to type in my vitals. In the electronic wilderness, I’m the prey and she is about to cut for sign. I spell my last name for Brown, and while she slowly pecks in the letters she explains that a typical investigation at FinCEN starts with a computerized check of commercial credit bureaus, because they keep accurate track of current personal data like names and addresses.
About 40 seconds later, step one in the investigation appears. It is a (largely blank) page listing which credit cards I use and showing their status. Nothing too remarkable here: any credit card company, loan officer, or car dealer can get this same information about as fast, and in fact will receive more specific information than commercial inquirers. The Fair Credit Reporting Act limits government access to credit records, eliminating details about what I have spent, where, and when. FinCEN, like other law enforcement agencies, has to get a subpoena if it wants to see individual records of what I bought at the mall last month. The meager sheet on me seems slightly forlorn and comforting at the same time.
“When we get these printouts,” Brown says, “the investigation basically confirms that there is a record on you.” That discovery doesn’t seem particularly spectacular—after all, who doesn’t cast a shadow in the digital world nowadays? But it’s mighty practical from a law enforcement perspective. When the phone rings at FinCEN, it may be the U.S. Marshal’s Service in a big hurry. If the marshals are pursuing a fugitive, someone they hope to catch by tonight, the basic information here—my home address—may be all they need.
But FinCEN has bigger muscles than just that. In 1994, the agency received about 6,000 requests for information from law enforcement groups already investigating someone. This support work is the largest component of FinCEN’s daily labor, and the Op Center typically assembles detailed dossiers on suspects who have already been identified. The dossier may be as basic as a few addresses, or it may extend into a hundred-page documentary on the entire life of a suspect—his businesses, family connections, driving record, even how many times he has been divorced.
Around the bend in the “L” from Brown, Larry Winnigar, an intelligence and research specialist, takes over the hunt for my spoor.
Using the basics established by the credit check—name and address—Winnigar turns to the realm of commercial databases. FinCEN has access to more than 30 of these, which combine information from public records in various states with national records compiled by marketing firms. I’m in here somewhere. So is Winnigar. So are you.
With a click of the mouse, Winnigar spins up the CD-ROM from the largest of these services, called Autotrac. A list of states appears in a box on the left, and Winnigar highlights the District of Columbia. Immediately a box on the right lists all the data available for D.C., by type. Winnigar highlights the listing for addresses, and in a moment I am looking at a list of 12 people who share my name in Washington. After highlighting the proper entry for me, Winnigar clicks twice more on the mouse and retrieves my Social Security number and its date of issue.
He grabs a pull-down menu, fiddles for a second, and the next thing I see on-screen is a list of all my neighbors in my Adams Morgan apartment building. I get a faintly dirty thrill from this, like a digital Peeping Tom. Here are the people I see in my hallway each day, laid bare on a computer screen in Tysons Corner. The mere possibility of peering into their lives seems to give me an almost automatic urge to do so. What about that bothersome woman one flight down? Does she have some secret that could be mine? What about the mysterious woman right next door, who never leaves her apartment? What can the machine tell me?
I don’t think my digital prurience sets me apart from most people. The urge to know is what makes us human. The very existence of the information draws all of us to investigate, to snoop. We are a nosy species, and the computer, by its nature as a keeper of information, is a tool of unsurpassed elegance and stealth.
I can see that several of the names on the list are out of date, but not by much. Because there are 28 entries at the same address, Winnigar doesn’t have to be Sherlock Holmes to deduce that I live in an apartment building. If I were a big-time bank robber, the list of neighbors might be useful. Police could interview them about my comings and goings, or quietly clear them out of the way before busting down my door. God is in the details, of course, and one small bit of these bytes fascinates me: My letter carrier is listed. There is no name there, just a number indicating what postal route I am on. By checking with the post office, Winnigar explains, police could learn which carrier works that route, and then find and interview him, or even intercept my mail. Such steps are not the work of FinCEN; the agency simply sets the process in motion by digging up the information. But it all starts with knowing where to look, and FinCEN knows where to look.
As he would when profiling a typical suspect, Winnigar next expands the search by checking for relatives. Most criminals are remarkably stupid about hiding money, and if they don’t put it straight into their own account then they often give it to a relative to hold. Finding those relatives proves remarkably easy. Checking by surname, Winnigar quickly locates my brother in Virginia, and then my mother. Since illegal profits are often hidden in real estate, Winnigar takes only a minute more to locate my mother’s condominium in Florida. Tap, tap, and we even get a list of her neighbors there.
The key to all this tapping is the mountain of data assembled by state governments. The CD-ROM is simply an assemblage of public records from around the country. Winnigar “steps back” to the main menu to highlight several of the states that offer the most comprehensive records, and therefore the richest datafields to dig in. Oregon is a good one, listing not only the routine stuff like driver’s licenses and vehicle registrations, but also boat registrations, department of corrections records, doctor’s licenses, and lien files. In quick succession, Winnigar summons a similar list of resources for Nevada, New Hampshire, New York, Texas, and Washington state.
Twitch the mouse a few centimeters and you enter Florida, the data hunter’s El Dorado. Because of of its well-deserved reputation as a drug dealer’s paradise, the state passed a number of “sunshine” laws in the 1980s that put an astonishing array of records in the public sphere. In theory, this gives criminals no place to hide their assets. The free market jumped in, and commercial transcribers entered everything they could find on CD-ROM.
Winnigar highlights “Florida,” and 15 databases appear in a box on the right. He picks driver’s licenses to start with, and begins assembling a profile of John Sheddan, a former deputy sheriff who happens to be a friend of someone at FinCEN and who allows himself to be digitally frisked in training exercises.
I have to wonder if Sheddan knows how thoroughly we are going through his pockets. The next 15 minutes convert my Peeping Tom lust into vague embarrassment, as though I were running my hands over this stranger: Along with height, weight, address, and phone number, we quickly learn that Sheddan has a concealed-weapons permit, drives a 1983 Mercedes Benz (black, 5 cylinders), and has had five traffic accidents in recent years (I scanned the police reports on all five). Along with his current and previous addresses, we see that Sheddan was divorced in 1986 after a marriage that lasted only 23 months. His ex-wife and all cars registered to her are close at hand.
There are no “hits” for Sheddan in some of the other databases—he has no license for selling salt-water products (drug dealers in Florida often conceal their real imports by setting up false commercial fishing operations), nor any outstanding liens on his property. Nor has he ever applied for workers’ compensation in Florida. He isn’t wanted in Broward County, has no known criminal history or fictitious names, is not a corporate officer, and owns no boats. He is not licensed to sell real estate.
The dossier, based entirely on public information, seems lewdly intimate. In 15 minutes I have the life of a total stranger detailed before me, down to the length of his marriage.
The search into my own personal background could go deeper if FinCEN dipped into the more sacred records of law enforcement agencies. Winnigar could check things like my citizenship and the dates of my travel overseas by calling the Customs Service, for example—but he declines to do so.
“We don’t surf the Net,” he says, explaining that by law FinCEN analysts must have a demonstrable suspicion of criminal activity before tapping into the law enforcement records. By law, the agency on the other end of the transaction—say, the FBI—cannot release any information to FinCEN without a formal statement that an investigation is in progress. By law, that restriction is enforced by internal guardians within the various law enforcement agencies. By law, watchmen watch themselves.
The purpose of such restrictions, Winnigar notes, is to “prevent Big Brother from saying, ‘I don’t like Patrick Symmes, what have you got on him?’”
There it is again, the dark icon that keeps cropping up in conversations about FinCEN: “Big Brother.” In fact, the phrase crops up within FinCEN fairly often, to the dismay of some and the evident amusement of others. When one analyst makes a joke about Orwellian snooping, FinCEN’s media director lets out an image-conscious groan. “Oh, there you go again with that Big Brother stuff,” she says.
Timothy McVeigh didn’t find it funny, either. Right from FinCEN’s inception, the vigorous fringe of gun crazies, black-helicopter spotters, and minutemen has had the obscure agency in its rhetorical gun sights. FinCEN has been described in militia literature as the enforcer of the dread New World Order and a power base of a shadow government.
In one of his less guarded moments, McVeigh explained to friends that “the government” had installed a microchip in his buttock. This might strike you and me as a bit blotto, but to loonies raised on conspiracy theories and fed a steady diet of The X-Files it makes perfect sense. While most of us were still picking our noses and trying to decide if the Cold War was over, the militia boys had shifted to the next paradigm. A thick report called “FinCEN and the New World Order” has been distributed at militia bake-offs and book sales almost since the agency came to life. Among the daft pronouncements on Masonic conspiracies and bar codes as the “mark of the beast,” FinCEN is alleged to be tracking militia members by installing tiny transceivers under their skin. The packet even includes photocopied ads for transceivers that are actually made and marketed today. As the ads show, however, these devices, each the size of a grain of rice, are used for inventory control of cattle, and can only be read by a handheld receiver within a few feet. They might help you keep track of whether old Bessie has passed through the feed chute yet, but they can’t be used to track people from outer space. McVeigh and his swamp-fever buddies are nuts if they think the government is plugging into people’s butts. Or maybe just ahead of the curve.
“Actually,” says Stan Morris, the director of FinCEN, “our main plan is to put a Pentium chip in everyone’s head.”
This is a regular shtick for Morris, one he plays extra-dry. A wisecracking, voluble veteran of various fiefdoms in law enforcement—he led the U.S. Marshal’s Service and worked for the federal drug czar—Morris has a broad, harmless look to him, so that he resembles a big brother more than a Big Brother. His favorite reply to the conspiracy theorists is to immediately confess FinCEN’s role in microchips, shadow governments, and even an invasion of America.
“We have been alleged to be the intelligence arm of a United Nations army that is going to suck the very juices of America’s freedom,” Morris says. “Obviously, my mother would not be happy to hear that.”
But once the fringe players have been dismissed with a wisecrack and a flick of the wrist, Morris is still confronted with two sets of critics who are not so easily turned aside—other federal agencies and longtime privacy advocates.
Inside law enforcement, there are plenty of people who simply don’t like FinCEN—or Morris. Their criticisms stem not from concern about Orwellian snooping but from the usual Washington disputes over turf, utility, and bruised egos. Both the FBI and the Customs Service have downplayed the importance of FinCEN and are developing their own systems for profiling suspects and searching financial databases. In off-the-record conversations, other white-collar cops describe Morris as a canny bureaucratic infighter, a survivor with sharp elbows and a love of publicity. Morris returns the favor by turning off my tape recorder to tell off-the-record stories about the incompetence of rival agencies.
Rivalry alone, however, does not account for the kind of criticism leveled by one Treasury Department official who is an end-user of “proactive” analysis generated by the AI system. This official—who chose anonymity for fear of antagonizing Morris—says that FinCEN is so slow preparing its reports that the information is “generally pretty worthless” by the time it arrives.
That kind of dissatisfaction was apparently behind a decision last year by the Justice Department to eliminate a small subsidy payment it had made annually to FinCEN. After years of steadily increasing FinCEN’s budget, Congress has now held the agency at its current annual level of about $22 million while, for almost the first time, raising questions about the agency. Sen. Al D’Amato of New York, chairman of the Banking Committee, recently asked FinCEN to explain exactly what its purpose is—whether the agency planned to continue primarily as a support service for other law enforcement groups or was attempting to become an “operational” agency launching its own investigations.
The second group of critics are those Morris calls “the civil rights people.” Morris passionately defends his agency’s handling of legal rights and privacy concerns, describing various internal safeguards to prevent the misuse of information. Anyone caught making an unauthorized search into someone’s private records would be fired, Morris says. “Obviously there are concerns about privacy, and we respect that.” The specter of “data abuse” actually wakes him up at night, he claims.
The critics are more concerned that FinCEN’s internal guardians could doze off during the day.
“I’d say the jury is still out on FinCEN,” says Charles Intriago, editor of the Miami-based newsletter Money Laundering Alert. Intriago believes that the little agency is having a huge and positive effect in the battle against money laundering, but fear clouds his enthusiasm. If the data—and the technology for exploiting it—are ever turned toward political ends, he says, “God knows what could happen. Can you imagine how their databases would have been used in the Nixon administration?”
Nixon’s habit of tapping the IRS to harass his enemies looks quaint when compared with the damage a multimedia Machiavelli could do today. Aside from its access to secrets of the financial world, FinCEN does conduct “data sharing” with not just federal and state law enforcement agencies, but also intelligence agencies. FinCEN won’t name the agencies with which it shares data (CIA? NSA? DIA?), won’t say which way the data is flowing (in other words, who is helping whom?), and what kinds of data go back and forth (Does the CIA share its profiles of foreigners? Does FinCEN check the financial records of suspected spies?).
Lots of questions. An afternoon in the Op Center answered few of them. Under the Freedom of Information Act, I had applied for my own FBI file a year ago and was still waiting for a response. The prospect of having FinCEN’s Winnigar just surf over to some secret “email@example.com” website and plug my name in was tantalizing, but Winnigar immediately declined. No surfing the Net. His resolve in the face of my pleading was hardly comforting, since nobody likes reporters anyway. But what if the plea came from a fellow cop in need of a little break to close a case, not a nosy reporter? What if the request were in a “good cause” rather than just for an article? What is to stop someone?
There are laws, and internal regulations at FinCEN, and watchdogs set over the staff to see that no one is tapping where they shouldn’t. But in terms of outside regulation of the agency, there are few options. As with the FBI and other keepers of secrets, FinCEN is supposed to be held within the law by a combination of internal enforcement and the weak pressures of the larger society—a free press, whistle-blowers, and an ever-vigilant Congress. The feeble nature of such a policy is only emphasized by the fact that FinCEN has attempted to set up an advisory committee made up of outside privacy advocates, including a member of EPIC. Formed in the last few months, the committee has had one meeting at which it was discussed how to form the committee.
The staff at FinCEN is certainly aware of the potential for abusing the oceans of data at its fingertips. Aside from the quiet jokes about Big Brother, there is a kind of reverential hush during Winnigar’s demonstration of all the various things FinCEN can find out about me. I’m glued to my chair in the Op Center, watching Winnigar flick through my life and feeling slightly paranoid about my mailman, when another FinCEN analyst standing nearby emits a low whistle.
“It’s amazing the things you’ll find in these databases,” he says quietly. “It’s almost scary.”
If profiles based on public information are enough to scare you, then Room 2431 offers some real cyberchills. Windowless, not much bigger than a typical janitor’s closet, the room is decked with Sun workstations and oversize monitors. Through this cramped space pass hundreds of millions of pieces of information a year.
This is where FinCEN goes operational. The support work done in the Op Center relies largely on public records. But the artificial intelligence center in 2431 is a far more secretive place, where actual investigations are launched. This is where FinCEN computers sift and collate the vast stock of financial data sired by the Bank Secrecy Act of 1974. The act requires banks and other financial institutions—check cashers, currency houses, auto dealerships, wire transfer shops, even casinos—to report every cash transaction of $10,000 or more to the IRS. Travelers entering the country with that much cash complete a similar customs form.
About 11 million of these reports were filed with the IRS in 1994. A majority of the paperwork is literally that—paper forms like the Cash Transaction Report (CTR), which resembles a 1099 tax sheet. Bank employees complete the CTR by hand, then send it to the IRS’s Detroit Computing Center, where the information is laboriously keyed into IRS computers and transmitted to FinCEN. The whole process can take up to six weeks; FinCEN, with the eager support of banks, is replacing paper with pixels. One-third of CTRs are now filed electronically; the data stream is faster and, because FinCEN is eliminating certain routine transactions from the reporting requirement, the stream is gradually narrowing.
To dip into that faster flow of numbers, FinCEN employs its AI software—a fancy name for a rule-based program that mimics the questions a human expert would ask. The computers in Room 2431 can then sift through the entries on each CTR, from last names and occupations to account numbers and citizenship. There are about 150 boxes, known as “data fields” in computer jargon, that a teller must complete on each form. Multiply that by the hundreds of thousands of forms entering the system every month and it becomes a picture of the financial doings of a nation, a picture that FinCEN studies for the hidden patterns of money laundering.
Despite the computer gizmos and analytical programming, the best way to spot money laundering, however, is still plain old human suspicion. Bank tellers are required to complete a special form for any transaction they deem “suspicious.” Deposits over $10,000 can be labeled suspicious if they are too frequent or are inexplicable by normal business operation. Even deposits under $10,000 can be reported if they look like “smurfing,” a term for breaking large cash sums into smaller amounts that evade the reporting limit—$5,000 a day for three days, instead of one $15,000 deposit. The form indicating that someone felt a suspicious transaction was under way is, of course, a very loud warning bell to the AI system.
“We are teaching the computer to react, to behave rationally, and to make intelligent determinations in much the way an expert would,” says Matt (who requested that his last name not be used), an occupant of 2431 and one of the analysts who has programmed hundreds of rules into the AI system.
The convergence of an increasingly digitized world with an agency that knows how to decode the blips means that damning evidence can be rendered in exquisite detail. During my first visit to FinCEN early in 1995, Matt showed me a case he was working on. The visual display was undramatic: a spreadsheet showing the bank records of three businesses and naming the five men who operated them. The case involved a group of greengrocers in the New York area who appeared to be involved in a creative scheme to bilk the Department of Agriculture with fraudulent food-stamp claims. The conspirators illegally bought food stamps from the poor for pennies on the dollar, then redeemed them at face value using a maze of bank accounts and different businesses.
Now, eight months later, Matt clicks a few times on the mouse and summons the data reaped from further digging into the bank transactions of the Brooklyn greengrocers. The spreadsheet appears, a white box on a blue background. Matt was able to trace the bank accounts to new individuals, who in turn led to new businesses that were participating in the fraud. Now the left side of the chart shows five groceries and the seven men who own them. Across the top, columns break down current financial transactions: total cash withdrawn this year, total last month, number of transactions over $10,000, portion of withdrawals that were in $100 bills, and “#STrans,” indicating what portion of the transactions were labeled suspicious by bank tellers—in this case, a large portion. Absent from the spreadsheet is any evidence of large cash deposits. Unlike a normal grocery store, which earns mostly cash every day, these business apparently earned almost no actual money.
“A strange way to do business,” Matt notes with a smile.
Some combination of the figures on Matt’s screen—the number of suspicious transactions, the abnormal cash flow pattern, the percentage of hundred-dollar bills—had triggered FinCEN’s computers, sending up a warning flag for the analysts. (To protect its methods of investigation, FinCEN would not reveal the specific analytical rules that drive the system.)
And it is precisely the automatic, computer-driven nature of the AI system that worries outsiders who fear that FinCEN could become a sort of digital Star Chamber, prowling through the ranks of the innocent rather than focusing on actual evidence of wrongdoing.
“It starts to use the technological capability that exists to really change our notion of what a law enforcement investigation is,” says David Sobel, a staffer at EPIC who sits on FinCEN’s unpaid committee of outside privacy advisers. “In the past you started with some indication of criminal activity and then investigated. The capability is now starting to appear where they can collect huge amounts of information, analyze it, and only then see if there is indication of criminal acts. It really turns the process around.”
But FinCEN’s Matt says that the agency’s man/machine attack on financial skulduggery is focused only on bad guys—or “legitimate targets” in the jargon. Matt insists that searches are never random and that the AI system does not simply cough up a daily list of people to investigate. Yet in brochures touting the agency’s abilities, FinCEN proudly describes the AI system as “proactive,” searching out suspicious patterns that human intelligence might never notice.
It comes down to an issue of how broad a search should be. The humans running the system give the computers certain parameters, or “triggers,” which narrow the search. But not by much. Asked for an example of how the searches are limited, Matt replies with a recent example: “Jewelers in South Florida.”
In that case, the Dade County police initiated the search with a call to FinCEN. They did not ask for a search of one suspicious jeweler, but rather of jewelers in general. Merely being in the profession in South Florida was grounds for investigation. Matt then plugged in the words “jeweler” and “jewelry shop,” either of which could be listed under “occupation” on CTR forms. Then he cross-referenced the relevant zip codes and turned the AI program loose to inspect the CTRs of all who fit those parameters.
The result is the last thing you expect to find in a government agency: art. With a click of his mouse, the result of a typical AI search appears on-screen above Matt. “We call it the ball of string,” he explains, a fitting description of the densely woven sphere of colored red, green, and blue lines. By clicking and dragging with his mouse, Matt zooms in on the yellow portion of the sphere. He has to zoom in twice more, expanding the view by orders of magnitude, before the tiny numbers along the rim become clear to the human eye. Each of the hundreds of entries visible in this small corner of the larger sphere represents a single bank transaction, and connects—via one or more long yellow lines arcing across the circle—to some entry on the far side, whether a bank account (brown), an individual (blue), or a location (green).
“The lines in between,” he says, “are showing how the people are connected together, that this is in fact one group.”
In essence, this is a standard “link analysis” used by police investigators all over the country for decades. But human investigators, pouring over file cabinets of paper documents, could take months to produce even the simplest diagram of related companies and bank accounts. FinCEN’s computers multiply the scope and speed of investigations. In less than an hour the AI system can produce a diagram so crammed with data that entries must be shrunk down until they can’t be read with the naked eye and the end product seems as tangled as a ball of string.
It is a visual metaphor for the information age itself: fantastically complex, incomprehensible to the naked eye, full of invisible messages, both frightening and awesomely beautiful.
I’m not in the habit of shifting large amounts of cash around the country, so Matt was unable to locate any records of me in his database. And I never made it into anybody’s ball of string. Most Americans are in the same position—we don’t do anything that would attract the attention of the computers in the Op Center or Room 2431, and questions about FinCEN, civil rights, privacy, digital snooping, and the proper role for government in the information age remain an abstraction.
But FinCEN has had a serious, butt-kicking effect on the lives of plenty of people. A few months after my session with Matt, 43 men were indicted in the New York area in the food-stamp scam. The indictment alleged that the ring drew as much much as $100,000 a month from the food-stamp program, for a total of more than $3 million in fraudulent claims.
Clearly, FinCEN does its job well, catches criminals, and avoids interfering in the lives of millions of us ordinary citizens. The staff is trained to avoid abusing privacy, and the rules are apparently enforced, since no one inside or outside the agency has yet detected a single case of improper investigation.
But that isn’t enough. You don’t have to be a militia member to realize that the technology of the information age has created unprecedented means of invading our privacy. FinCEN may be acknowledging at least the possibility of such abuse by recruiting critics like Sobel into review panels, but given the certainty of human foolishness and sufficient time, the possible becomes the inevitable, if not at FinCEN then somewhere else in the federal government. The FBI, Customs, and DEA run their own, similar programs of profiling and are developing the muscles of their computers to do the same kind of aggressive inspection of our lives.
It is hard not to marvel at all the bells and whistles in FinCEN’s domain. Yet amid the reams of data, the easy mastery of secrets, and the fine cogitations of artificial intelligence, there are some less obvious lessons.
One is about how little we know about what is known. There was that curious little ritual at the entrance to the Op Center, where the FinCEN staffers stopped to rub their identification cards against the white, wall-mounted box, leaving a magnetic signpost of their entry. Such small steps are part of our life now—plugging into cash machines and keying entrance codes to buildings, paying with credit cards and leaving vast electronic wakes behind us. This omnipresent electronic environment comes with its own, almost subconscious anxieties: Unless you are a true techie, you don’t have any idea how the magnetic box works, or what information has just been transferred. Some hidden machine has just passed judgment on you—but what judgment? Does it know your name specifically, or merely that “you” are a member of that protected class of people whose cards carry the correct trace of code? Did it note the time and location of your entry? Can that data be used to reconstruct your movements? It isn’t hard to believe that a government interested in the ingress and egress of dollars in my bank account might begin to take an interest in where the rest of me is coming and going.
The spiral of questions can easily spiral out of control. We live in a womb of such electronic messages, recordings, and notations, an environment that is quite naturally invisible, a parallel universe that also, quite naturally, induces a certain paranoia. I kept thinking back to that white box on the wall, and asking myself, who knows what is known anymore?
Against such frightening musings I had the comforting memory of what happened that first day when I left the Op Center. Winnigar had just finished his dazzling drive through my digital life, dishing up the names of family, locations of homes, licensing of cars, mail routes and dozens of other aspects of my life. At the end of 15 minutes, slightly dazed by what I had seen, I waited quietly while he made a printout of the entire dossier for me to take home, a souvenir of the kinds of information that are in government’s—in everyone’s—hands.
After a minute I realized Winnigar was still watching the printer. Nothing happened. He tapped his keyboard a few more times, and then wandered over to the machine. A few people joined him. Pretty soon the cream of the government’s computer crop were poking the printer, wiggling the cables, fiddling with the switches, tapping the keyboard, and scratching their heads. Just a minute and we’ll have it, they explained. The minutes added up.
In the end, I never did get my dossier, but I was slightly cheered to discover that Big Brother is subject to the same bugs as the rest of us. For weeks I kept smiling at the thought of the omniscient government choking on a simple printout. Meager consolation in the age of information, but it may be all the consolation we get.CP
Art accompanying story in the printed newspaper is not available in this archive: Darrow Montgomery.