About
LLEA develops and evaluates LLM-based methods for law enforcement data analysis. The project is structured around three core pillars:
- Transcription & Translation
- Heterogeneous Data Analysis
- Fraud Platform Detection
The project is carried out in close collaboration with four Austrian agencies — BMI, BMF, FMA, and ZCB — ensuring methods are grounded in real operational needs.
Goals
Regulatory, Ethical & Social Analysis Regulatory (e.g. EU AI Act), ethical, and social frameworks are comprehensively analysed and assessed to evaluate the use of LLM-based methods in existing processes of information gathering and decision-making.
Development of LLM-Based Methods LLM-based methods are developed for each pillar, either building on pretrained models or adapting them to specific application contexts through targeted fine-tuning.
Systematic Validation The developed methods are integrated into a proof-of-concept environment and systematically validated against predefined criteria in close collaboration with the agency partners.
Training & Knowledge Transfer Training courses based on the approaches developed in the project are established and permanently integrated into the continuing education curricula of the participating agencies.
Funding
Funded by the Austrian Security Research Programme KIRAS, financed by the Federal Ministry of Finance (BMF) and managed by FFG.