Amid increasing data volumes and evolving fraud schemes, federal agencies are recognizing the potential of artificial intelligence (AI) tools to enhance oversight and counter fraudulent activities. The Pandemic Response Accountability Committee (PRAC), responsible for overseeing pandemic-related spending, has identified AI as a valuable tool to identify potential fraud in pandemic spending data. With AI’s ability to quickly analyze vast data, agencies are exploring its potential to flag fraudulent activities and protect federal benefits programs.
AI’s role in identifying fraudulent activities
The PRAC Chief Data Officer, Brien Lorenze, highlighted AI’s significance in addressing fraud within pandemic spending. By harnessing AI’s capabilities, agencies can identify synthetic identities and shell corporations that fraudsters may exploit to access federal benefits improperly. Lorenze emphasized the “target-rich environment” for AI in pandemic oversight and its potential to expedite the identification of fraudulent patterns.
The PRAC holds approximately 95 agency data use agreements and manages around a billion transaction records from eight agencies. Navigating this vast dataset poses challenges due to varying rules and governance over data usage. Ensuring ethical use and accuracy of AI tools remains a priority for the Biden administration.
AI’s potential to combat evolving fraud
While the Biden administration works towards ethical AI implementation, agencies grapple with the increasing threat of fraud schemes accelerated by malicious large-language models like FraudGPT and WormGPT. These models enable fraudsters to create malware and phishing attacks more efficiently, posing new challenges to agencies combatting fraud.
Lorenze acknowledged that tools like FraudGPT and WormGPT can aid fraudsters in crafting more sophisticated attacks, underlining the need for agencies to understand these tools to counteract them effectively. AI’s duality requires agencies to grasp its potential to enable fraud and its capacity to identify fraudulent activities.
Enhancing intelligence with AI tools
Heather Martin from the National Geospatial-Intelligence Agency’s Directorate for Data and Digital Innovation highlighted AI’s potential in aiding analysts to navigate vast volumes of satellite imagery. As the volume of data grows, AI tools help intelligence analysts efficiently locate objects, enabling them to focus on higher-priority tasks and in-depth analysis.
While AI holds promise, agencies like U.S. Citizenship and Immigration Services (USCIS) are taking cautious steps towards its adoption. Damian Kostiuk, Deputy Chief Data Officer for USCIS, emphasized that initial use cases of AI are relatively mundane due to legal considerations. USCIS uses AI to pre-draft documents, but human oversight remains crucial to ensure accuracy and regulatory compliance.
Challenges in data environment and sharing
Kostiuk highlighted that AI requires extensive data for training, yet many federal agencies operate in data silos, hindering the availability of comprehensive datasets for AI models. Sharing data across agencies remains a challenge, limiting the effectiveness of AI technology.
The Commerce Department is conducting an agency-wide data maturity assessment report to tackle these data-related challenges. This assessment aims to evaluate how agencies govern, manage, and leverage data, identifying gaps and opportunities for improvement. By enhancing data quality, governance, accessibility, and sharing, agencies can harness AI’s potential more effectively.
The evolving landscape of fraud, propelled by advanced language models, necessitates agencies’ use of AI to comprehend and counteract fraudulent activities. The PRAC’s focus on leveraging AI to identify potential fraud in pandemic spending exemplifies the growing recognition of AI’s potential to streamline oversight and enhance fraud detection capabilities. As agencies tread cautiously, addressing data challenges and improving data utilization will be essential to unlock AI’s true potential in combating fraud and bolstering oversight.
