Can AI Help Banks Handle Venezuelan Sanctions?

The Trump administration’s tightened sanctions on the regime of Venezuela’s Nicolás Maduro have tested U.S. banks’ transaction monitoring systems, with some institutions responding by refusing to accept payments to and from the country, a move that could worsen the country’s humanitarian crisis.

But artificial intelligence may help banks better discern good money from bad.

The list of Venezuelan-government-connected companies with which U.S. banks are forbidden to do business has been steadily growing along with pressure on financial institutions to comply. National Security Adviser John Bolton took to Twitter last month about the issue, warning bankers not to help Maduro or his allies.

This is a pressing issue for banks in south Florida, according to Daniel Gutierrez, vice president and regulatory risk manager at the $3.9 billion-asset Ocean Bank in Miami. South Florida has the largest Venezuelan expatriate population and many Florida banks do business with Venezuelan companies and immigrants.

“It’s not a simple issue because you’re dealing with red flags that deal with corruption,” he said. “When I say corruption, I’m talking about sophisticated means to [game] financial institutions’ ability to find this.”

Venezuela has been in turmoil for several years as Maduro took power in 2013 and sought to consolidate his hold on the country. Though Maduro was ostensibly reelected last May, the U.S. and more than 50 other countries believe the election was rigged and have recognized Maduro’s opponent, Juan Guaidó, as the country’s rightful leader. It has led to escalating sanctions by the U.S. against Maduro’s regime as massive protests have spread throughout the country in opposition to the strongman’s rule.

But sanctioned companies are doing everything they can to avoid being cut off by routing transactions through third parties and shell companies in other jurisdictions, said David Schwartz, president and CEO of the Florida International Bankers Association.

“A lot of these roundabout transactions don’t necessarily come from Venezuela, they’ll be routed through let’s say Panama, so getting to the origin of the funds is really important,” said Ivan Garces, practice leader, risk advisory services at Kaufman Rossin. “Sometimes you see payments coming through third parties and you have to ask, what was the origination of the funds? It’s getting to the point where banks are scrutinizing every transaction to determine a few things: Is it consistent with this customer’s business as I’ve known it, can I get comfortable with the source of the funds and can I tie it to the customer’s legitimate business?”

Garces said banks are applying enhanced due diligence to their Venezuelan customers in an effort to better understand them and the type of business they do, including whether or not they transact with the Venezuelan government. In some cases they’re forcing customers to document their relationship with the government or lack thereof.

“Each transaction requires a very careful review that goes beyond what we’ve been accustomed to doing,” Garces said.

Like most banks, Ocean Bank has suspicious activity monitoring software to prevent money laundering but those that want to circumvent the sanctions appear well-versed in how the system works. “If a $2 million wire hits my bank, [the software] will alert me and prompt me to ask for more information,” Gutierrez said.

The bank will ask for supporting documents like invoices, contracts or bills of lading that appear legitimate.

“They have signatures and notarizations and at face value [they] are legit,” Gutierrez said. “They have access to notary publics, they know what we are going to require to be comfortable with the transaction.”

Another challenge is that the sanctions are written broadly.

“This has created fear on the part of financial institutions because they have to be extremely careful not slip up in terms of detecting transactions that come through their institution that are related to anyone on the list or any related parties to them,” Schwartz said.

For example, one sanction prohibits the exportation of crude oil diluents from the U.S. to Venezuela.

“I’m a banker. What is a diluent?” Gutierrez said. “Now I know it is a chemical substance used to dilute the crude petroleum or crude oil. If a wire comes into my bank, how am I going to know if I’m negotiating or crediting a customer for a transaction that has to do with a diluent?”

The bank has to stop the wire and analyze it. If there might be something suspicious about it, the bank will get additional documentation from the customer. An attorney may be brought in to assess the legitimacy of the wire. This costs the bank time and money and the customer is affected.

“The customer might say, wait a second, where’s my money? I haven’t been credited,” Gutierrez said.

Some banks are just giving up and de-risking entirely, Schwartz said.

“We’re starting to see banks outright rejecting transactions because it’s much easier than trying to decipher the regulation and then determine if the transaction may or may not be related to somebody on that list,” Schwartz said. “Why run the risk of having to analyze the exception and know whether or not it fits within the exception? It’s easier and less risky to just say no and reject the transaction outright.”

Gutierrez said he has seen institutions return transactions with an address in Caracas when the benefit of running the transaction didn’t appear to merit the cost of due diligence around vetting the transaction.

“If a customer does not have a good standing relationship and is creating much more effort and work and risk than there is benefit to the institution, then why wouldn’t you consider perhaps not banking that customer?” he said. “This is a business ultimately.”

At Ocean Bank, however, “our goal is to avoid de-risking, and not just get rid of accounts because of concerns,” Gutierrez said.

The banks willing to do the extra due diligence required for these transactions are having to divert resources to this, Schwartz said.

Consumers are affected by the sanctions, too. There are many Venezuelan families sending remittances back to their home country “to try and help their families survive there given the situation,” Schwartz said. A couple of towns are almost all Venezuelan expatriates.

“So for us in South Florida, it’s important to try and work through it as best we can to try and keep those flows going and help these people,” Schwartz said. “There is suffering.”

Can AI help?

A crop of new, AI-based anti-money-laundering programs tout a more intelligent, sophisticated approach to detecting suspicious transactions, with fuller vetting of companies and people and an ability to understand all the counterparties with which they are affiliated.

Providers include Quantaverse, Quantexa, Merlon, Zest Finance, Ayasdi, ThetaRay, IBM and Nice Actimize. Quantexa, for instance, says it can trace a transaction through any shell companies or partners that may be used to funnel the transfer of money.

Gutierrez says such technology can help, within limitations.

“People talk about artificial intelligence a lot,” he said. “However, as with any other technology, you cannot just purchase the artificial intelligence, take it out of the box and plug it into your bank systems and expect it to do marvels. You have to make sure this intelligence is going to help you, and you have data integrity and sufficient and clean data for the system to learn from itself.”

Banks also have to make sure the product makes sense for them, as the vendor is going to assure them it can do anything. Gutierrez also noted that he doesn’t know of any system that can read contracts in Spanish.

“I know I can feed a system with the 800-plus state-owned entities that there are in Venezuela,” he said. “And I know that by feeding those names the system can go ahead and detect them.”

But there are still gray areas that call for human analysis, he said.

Schwartz also said he has seen vendors with interesting platforms.

“A machine can analyze patterns of behavior, patterns of transactions and draw from a variety of sources to try and obtain information that might help break through some of these nontransparent transactions,” he said.

The problem is that these sanctions are very complicated.

“It’s not just a question of the person that you’re facing,” Schwartz said. “It’s a question of the transaction being traced all the way back to its origin. Knowing all of the parties that were involved and being able to determine if they’re a front person or if they’re not a front person can be a challenge. It’s not a perfect technology.”

AI-based AML software can reduce false positives, Schwartz said. “But it doesn’t uncover all of the false negatives, the type of case where your normal monitoring systems have not caught the transaction,” he said. “It’s better than the systems that were being used. It’s still in kind of a testing phase in terms of how well can it can detect these types of parties that are trying to hide themselves and transactions. But it’s definitely progress over where we were five or even three years ago.”

Regulatory concerns may also be an issue. U.S. regulators have given institutions the green light to experiment with AI and other technology to fight money laundering, but privately many bankers complain that examiners are still skeptical of new systems.

Garces also offers a mixed review of the effectiveness of the newer monitoring systems.

“There are opportunities for technology to play a role,” he said. “However, I don’t think technology alone at this stage is going to do it. Smart systems can be used to identify unusual patterns, but at the end of the day a judgment still has to be made as to whether the activity is fact unusual or not, tied to legitimate purpose or not.”