Please note that log in for BLAW products will be unavailable for scheduled maintenance on Sunday, February 5th from approximately 4 AM to 5 AM EST.
Bloomberg Law
Free Newsletter Sign Up
Bloomberg Law
Advanced Search Go
Free Newsletter Sign Up

States Deploy Artificial Intelligence to Stop Unemployment Fraud

Sept. 22, 2020, 10:30 AM

States are using artificial intelligence and other high-tech tools to identify and deny billions of dollars in fraudulent claims as they deal with a surge in Covid-19-related unemployment.

“These benefit programs are clearly very outmanned and outgunned,” said Jon Coss, CEO and founder of Pondera Solutions Inc., a firm that ferrets out fraud, waste, and abuse in health-care and government programs. “You can’t do this with just people, you have to have machines. The other guys do.”

Colorado recently developed a tech-based method to detect unemployment fraud it calls the “18th measure,” which is so effective it has helped identify more cases than the other 17 measures combined, said Cher Haavind, spokeswoman for the state’s Department of Labor and Employment.

Haavind said she couldn’t say more about the technique because the department doesn’t want criminals to know how the method is being used to stop them.

Nationwide, fraud is most prevalent in the Pandemic Unemployment Assistance program, which was created by the CARES Act and which gives states the option of extending unemployment compensation to independent contractors, gig workers, freelancers, and others who are ordinarily ineligible for unemployment benefits.

One disadvantage of the PUA program is that, given that it’s beneficiaries are self-employed, there are no companies through which states can verify pay—states must rely on self-attestation by the independent workers who file claims.

Fraud Widespread

As a result, many states are dealing with widespread PUA fraud.

  • California announced Saturday it would enact a two-week freeze on new unemployment insurance requests after Gov. Gavin Newsom’s (D) “strike team” issued a report with recommendations for how to better process claims and address a backlog of nearly 600,000 cases. The report noted that the state has seen a sharp spike in PUA claims, up to almost 120,000 a day on Aug. 31 and Sept. 1.
  • Arizona recently estimated that out of 1 million PUA claims, nearly 900,000 were found to be fraudulent.
  • Colorado announced earlier this month it halted $750 million to $1 billion in improper unemployment insurance payments from going out the door.

Other states believe they are victims of attack, but don’t know how many cases there might be, according to the California report.

More than three out of four claims made for assistance under the PUA in Colorado were fraudulent, Haavind said. From July 18 to Aug. 25, the department processed 62,498 filings, 48,206 of which were found to be fraudulent and 14,292 were deemed valid.

A common individual wouldn’t be able to accomplish the kind of sophisticated fraud states are seeing; rather, the surge in cases is a result of organized crime, with perpetrators based out of state or the country, Haavind said.

Fake Employees

“These scammers aren’t using this money for good,” Coss said. “It’s often used for guns or human trafficking. It goes overseas to subsidize terrorism.”

Scammers try to push through multiple claims from the same email domain or from the same postal address. They set up false companies with fake employees, in many cases using the stolen identities of people who are dead or in prison, Coss said.

“They buy false IDs off the dark web, identities of people in other states and countries. One claim involved an 8-year-old who reportedly made $100,000,” he said.

Another challenging aspect of halting PUA fraud is that “we’re talking about government benefits, not credit card fraud,” Coss said. “You could be holding up payments that are going to people who are depending on it to make a house payment or buy groceries.”

High-tech tools can help states to validate claims on the best front as best they can without slowing down claims, he said.

“By using machine learning and AI, you’re actually able to identify new and emerging trends, even predict them with some sort of accuracy and stop them before they go out the door,” he said. “It’s like Whac-A-Mole. AI helps identify what hole they’re going to come out of next, so you can be there waiting for them.”

To contact the reporter on this story: Tripp Baltz in Denver at

To contact the editors responsible for this story: Martha Mueller Neff at; Andrew Harris at; Tina May at