While some states have lost millions of dollars to pandemic fraud, others have turned to artificial intelligence to help process claims and catch criminals.
LOUISVILLE, Ky. — Fraudulent unemployment insurance claims, especially during the COVID-19 pandemic, cost several states hundreds of millions of dollars, prompting furious criticism from legislators and calls to improve and modernize identity management.
The Virginia Employment Commission said it lost around $470 million in unemployment payments, having identified around 258,000 potentially fraudulent claims for assistance during the COVID-19 pandemic. Kansas paid out up to $466 million in fraudulent COVID-19 benefits, and an April redacted pen testing report found its Department of Labor to be vulnerable to cyberattacks.
Similar problems in other states resulted in fraudulent claims worth as much as $80 billion submitted to the federal Paycheck Protection Program. The pandemic-related fraud was at least partly caused by governments quickly trying to get money out to help the most vulnerable, perhaps sacrificing oversight and burdening programs and systems not ready to cope with the level of demand.
“I would say it's a perfect storm,” said Jordan Burris, vice president for public sector strategy at identity verification company Socure, in an interview at the National Association of State Chief Information Officers’ annual conference. “Government agencies have always been on a maturation journey of sorts, and the pandemic forced an acceleration of what many of those plans were for a number of CIOs and digital service teams.”
In a bid to fight back against the fraudsters, some states turned to artificial intelligence and machine learning. Utah was doing so even during the pandemic, using AI to flag suspicious applications and send them for manual review.
In a paper released last year by NASCIO, Utah’s Director of IT Services Michael Hussey recalled being contacted by human resources and asked if he had resigned as they had received an unemployment application in his name from fraudsters.
“I was kicked out because of some technology that we implemented to detect fraud at that level,” Hussey said in the paper. “Luckily, the state caught that.”
Burris said AI can pull together someone’s personal identifiable information, the geolocation and IP address of where they are entering their information and verifiable documents like their passport or driver’s license. With that data ingested, the technology then can assign someone’s application a risk score, which rates how likely their application is to be truthful, fraudulent or needing a manual review.
There is much work to be done to refine AI and algorithms, including ensuring they do not perpetuate racial bias, considering those with a limited personal information footprint and by being more transparent about how the technology works, Burris said the technology could be more widely adopted if agencies are forward-thinking.
AI is “a current state,” as it is currently available to business and government leaders, he said. “However, in the public sector, it may be seen as more of a future state, and it's only because they haven't had widespread adoption of the benefits associated with it,” he said. “Also, to be clear, AI in and of itself isn't a panacea.”
Some in state government have been reluctant to embrace AI and other emerging technologies for fear that it will result in job losses and staff being forced out of their roles as agencies become less reliant on human labor. Jack Shaw, an innovation and technology strategist, said during a keynote address at the NASCIO conference that while that view is “understandable,” the future of AI in government is more nuanced.
"There is good news about AI, and that is that AI will not replace government decision-makers at any level,” he said. “The bad news for some of them … is that government decision-makers at every level who know how to use AI-enabled tools and technologies will replace those who don't."
Burris said governments should look to industries that deal with identity verification like the financial sector for insights on how to avoid similar fraud in the future and to ensure data is not compromised. The priority for governments looking to combat fraudulent claims should be to embrace new technologies and new verification methods, he said, rather than locking down systems and making access difficult.
Identity management does not have to be so stringent that it is “the barrier for someone getting access to government services and benefits,” he said. State governments and the private sector “can all act as a network and combat fraud together.”