Brown University eliminated 103 positions months after Alex Shieh, a sophomore at the time, used a combination of AI and public data to launch Bloat@Brown, a website that listed the names and titles of each of the school’s nearly 4,000 administrators. Shieh has since left the school to launch the Antifraud Company, which intends to use AI to identify the billions of dollars of outright fraud that is perpetrated against American taxpayers every year.
Although the topic of waste, fraud, and abuse hasn’t gotten as much attention since the flash in the pan that was the Department of Government Efficiency (DOGE), it remains a huge drain of taxpayer dollars.
In 2024, the Government Accountability Office (GAO) estimated that the federal government loses between $233 billion and $521 billion a year to fraud (based on data from FY 2018 to FY 2022), which the GAO defines as “obtaining a thing of value through willful misrepresentation” including “making material false statements of fact based on actual knowledge, deliberate ignorance, or reckless disregard of falsity.” Instances of this fraud include a physician billing health care benefit programs, such as Medicare and Medicaid, for “unnecessary, expensive testing on addiction treatment facilities patients” to the tune of $127 million, and an owner misrepresenting his finance company to receive “small business loan lender status and fees” for $71.7 million.
The federal government has been able to recoup some of this money—the Council of Inspectors General on Integrity and Efficiency recovered between $6.6 billion and $19.7 billion annually from FY 2018 to FY 2022—but this is less than 10 percent of GAO’s estimated yearly losses over this time period. Fortunately, the feds provide a strong incentive to uncover the rest of this monumental fraud: 30 percent of recovered funds.
The Antifraud Company believes this built-in bounty makes their Snitching-as-a-Service (SaaS) model viable for the long haul. The company’s unique SaaS model consists of three parts: “Forensic accounting, natural language processing, [and] traditional pound-the-pavement interviewing,” Sahaj Sharda, the company’s co-founder and CEO, tells Reason. He says misconduct always leaves a trace. To discover it, the Antifraud Company uses machine learning to parse through traditional market data and large language models to parse patents, regulations, and contracts. In addition, the company “go[es] out and talk[s] to people who are affected by these practices,” explains Sharda.
Shieh says that the company uses open source intelligence tactics similar to those used by digital investigation groups like Bellingcat. The Antifraud Company goes a step further by giving full-time investigative journalists access to the company’s “AI tools and engineers…so that they can do more than they could have ever done manually,” explains Shieh. Since its launch in June, the company has already uncovered $250 million in potential fraud.
