State of use of AI tax systems
As of October 2024, there are only a few publicly documented use of tax machine-learning algorithms in the Republic of Bulgaria.
What functions are performed with AI?
- Transaction Network Analysis (TNA): in 2019 the Bulgarian National Revenue Agency (BNRA) developed a predictive model to automatically identify missing traders and carousel fraud schemes. Reportedly, the model was developed based on a cooperation with the University of Exeter – Tax and Research Centre. The TNA system reduced the average time to detect missing traders by 15% and significantly increased the number of cases detected.
What data can be processed by these systems?
The taxpayer data used to train these models is not specified in publicly-available documents.
Are these systems regulated by specific norms?
Currently, the use of AI by the BNRA is not regulated by any ad hoc norm.
