State of use of AI tax systems
First reports of tax machine-learning algorithms used by the tax administration of the Republic of Latvia (SRS) date to 2019.
Systematic review of publicly-available sources reveals only a few AI systems initiatives deployed by the SRS.
What functions are performed with AI?
Based on publicly available data, two functions are performed with machine-learning algorithms:
- Taxpayer assistance: SRS deployed a chatbot ‘Tom’, in January 2020. Tom answers frequently asked questions in particular in five tax relevant domains namely, registration of a business, e-filing, annual submission of tax returns, eligible expenses and the eligible expenses mobile application.
- External risk-management (risk-scoring): A report of the EU Regional Cooperation Council indicates that Latvia uses ESCORT, a machine-learning algorithm which scores legal persons according to their predicted level of risk of fraud. ESCORT is used for a number of tax items, to detect and prevent a wide array of non-compliance risks. ESCORT is also used to detect and prevent fraudulent or undeclared labor.
What data can be processed by these systems?
The data used for the development and use of the VCA is not specified.
The data used for the ESCORT risk assessment are collected in one data warehouse and include the main information sources used for risk assessment in the SRS databases. ESCORT also process data from other governmental institutions such as the State Labour Inspectorate, particularly data on previously detect offenses such as undeclared labour or illegal migrant work.
The different features include: the average salary, hourly wage, the number of employees, of cash registers, of market places, the sector of activity, the number of offenses detected during controls and number of complaints received about envelope wages.
Are these systems regulated by specific norms?
The use of this model is not regulated by specific ad hoc legal norms.
