However worried you are about AI deepfakes, you're not worried enough

Writers are a good mark for fraudsters. We crave recognition, and we need money. I routinely get emails from people saying they want to help me sell my books. Sometimes they have some made-up name, sometimes they are impersonating real literary agents and editors. These cons are easy to spot, to me at least. But they must work on somebody, because they keep coming.

Online fraud is getting worse as technology is making it easier for criminals to perpetrate their scams. I was sent a report from a firm named Coveron that seemed excessive. Coveron sells scam-protection software and is a part of Nord Security. The report said that almost half of all Americans have been victims of online fraud.

I thought, that can't be right. That's too high. Almost half have fallen for a scam, and almost 80% have had an attempted scam, this report found. 

I went online to look for other data, and found a report from Pew Research last summer that seems to corroborate what Coveron argued. Pew found that 73% of all Americans have either been victimized or at least seen an attempt to defraud them. The most common theft was credit-card theft. People also reported buying fraudulent goods online, having their accounts hacked, and falling for phishing scams like the guys claiming they could sell 10,000 of my books.

To be sure, both reports are extrapolations. Coveron and Pew surveyed a number of people, and applied the results to the whole population. But whatever the real number is, it is unacceptably high. And I think it is going to get higher very soon.

An American Banker survey of high-level banking executives found that the institutions that were the furthest along in integrating AI into their systems were the most alarmed about AI-enabled fraud. Among banks that weren't using or only exploring AI, about 55% of respondents saw operations or third-party risk from AI as a high threat.

Among banks that were already deeply involved in AI, that percentage rose to 83%. 

That may be because artificial intelligence is increasingly becoming a tool criminals can employ to great effect. It is possible to use the technology to impersonate somebody, a so-called deepfake, and gain access to their accounts. This was possible three years ago, and the technology has only gotten more sophisticated and cheaper to access in the intervening time. This is becoming a palpable threat that banks cannot ignore, Julia Jakimenko wrote in a BankThink essay.

In the Netherlands, she explained, a man opened 46 bank accounts using deepfake technology to blend his own features with the features of people whose passports he'd gotten while pulling off a rental scam. That allowed him to get around facial-recognition software. This apparently was successful but very lucrative. The guy was being ordered to repay only about 6,000 euros to the bank he conned, and faced a 30-month prison sentence. 

Read more:

• Bitcoin's Overton window of opportunity is closing
• Maybe Wells Fargo's unions need a blockchain
• Iran built its own financial system, and has created a cyberfront in the war
• Fed meetings are about to get really awkward

It's not an isolated case. KYC-bypass tools are sold openly online. A fraud ring in Hong Kong pulled used the same techniques to open 30 fraudulent bank accounts. Jakimenko, the CEO of forensics company Cyberette, explains how new rules in Europe designed to combat deepfakes and AI-assisted fraud have loopholes that allow the kinds of verification-fraud scams I just described to proliferate. In the U.S., the Financial Crimes Enforcement Network has issued alerts, but there is no concrete method for stopping these attacks. Banks, and their tech vendors, essentially are left on their own to figure all this out. If you're responsible for cybersecurity at your bank, this is nightmare-fuel stuff. 

I wrote last month that defrauded consumers and consumer groups are trying to hold banks more responsible and liable for letting fraud occur on their systems. That is a complicated question that isn't new but these new AI-assisted crimes are going to turbocharge that debate, and it isn't likely to be just an industry debate. Of those banking executives we surveyed who were already deep into integrating AI, more than 80% expected the pace of new laws and federal regulations to pick up over the next three to five years.