A fraud-detection AI model trained on COVID-19 loan data could have alerted tens of billions of dollars in payments before they were paid, potentially easing the federal government’s clean-up burden, the U.S. government’s Pandemic Response Accountability Committee reported to Congress on Tuesday.
The truth is, of course, that the nearly 5 million Small Business Administration COVID-19 Economic Injury Disaster Loan applications used to train PRAC’s “fraud prevention engine” were only available after the program was already up and running, by which point billions of dollars had been paid out and later found to be potentially fraudulent.
Nevertheless, PRAC Executive Director Ken Dieffenbach told the House Oversight and Government Reform Committee during a hearing Tuesday that it’s time to use these lessons to prevent other cases of government misconduct, something the Trump administration and Republican Congressional leaders ostensibly value very much.
“It’s time to use this data to stop fraudulent schemes and hold wrongdoers accountable before taxpayer money is lost,” Dieffenbach told the committee.
According to the PRAC director, the anti-fraud engine was developed as a proof of concept to “determine whether such a model can be developed and to identify and address technical hurdles.” As Dieffenbach explains, the project appears to have been a success.
The engine is built with a number of modular components, including unsupervised machine learning models used to detect anomalies, supervised ML to identify patterns that match pandemic fraud incidents, and rule-based flags to detect invalid Social Security numbers and employer identification numbers. Dieffenbach said such small anomalies often identify hidden connections that suggest fraud, such as sharing bank account numbers between supposedly independent applicants.
And it can be done quickly. Dieffenbach said the AI can process 20,000 applications per second.
“Had our fraud prevention engine been in use in March 2020, it would have immediately alerted us to the potential tens of billions of dollars in payments for further scrutiny before the funds were disbursed,” Dieffenbach told the Oversight Committee.
PRAC says it has helped recover more than $500 million so far, but Dieffenbach said only a small portion of the payments could have been alerted to the engine before they were paid.
In other words, there’s a lot of work to be done, but only when it comes to pandemic fraud. But Dieffenbach hasn’t stopped the anti-fraud engine in PRAC’s coronavirus loan recovery efforts.
President Trump’s budget reconciliation bill, passed last summer, extended the PRAC’s mission to detect and prevent fraud in programs funded under the act, pushing back the commission’s termination date to September 30, 2034, and providing $88 million in funding to help expand its oversight work.
Since then, PRAC has begun working with inspectors general across multiple agencies to deploy anti-fraud engines and related analytics to assist in the oversight of programs funded by the Reconciliation Act. Dieffenbach said this effort has already helped identify specific programs where PRAC’s data and technology can provide valuable insights.
“We have identified several specific programs where our data and technology can provide valuable insights,” the director said.
Oversight Committee member Pete Sessions (R-Texas) says the engine needs to find a permanent home before PRAC disbands in eight years.
“A permanent solution that maintains the analytical capabilities and capabilities that have been built over the past six years is needed and needed,” Sessions said. “There are billions of records in that database, and it’s starting to pay for itself.”
We asked PRAC if there were any discussions taking place to find a permanent home for the engine, but we did not hear back. ®
