A graduate research project is raising questions about how longstanding transparency laws may be creating unintended risks for homeland security in the digital age.
Introduction to FOIA and AI Risks
A graduate research project by Melanie Simmons, a statistician at U.S. Immigration and Customs Enforcement (ICE), examines how the Freedom of Information Act (FOIA) interacts with modern data and AI capabilities, raising questions about unintended risks for homeland security.
Evolution of FOIA and Digital Data Challenges
Originally designed for a paper-based environment, FOIA now operates in a digital landscape where large volumes of data can be released, aggregated, and analyzed at scale, posing new challenges for transparency laws.
The 'Mosaic Theory' and its Exploitation
Simmons’ research introduces the 'Mosaic Theory,' highlighting how AI can combine seemingly unrelated data points from multiple FOIA disclosures to reveal sensitive information. This can be exploited by 'blind requesters' to reconstruct law enforcement-sensitive details or identify operational patterns.
Recommendations for Modernizing FOIA in the AI Era
The study suggests that existing data disclosure frameworks need to evolve. Simmons recommends modernizing FOIA processes to address data aggregation and inference risks, incorporating current federal data protection practices, and utilizing AI tools for pre-disclosure risk assessment.
