taxadmin.ai is a citizen-science open-access repository of the DigiTax Centre of Excellence, which compiles artificial intelligence (AI) algorithms used by tax authorities in the EU, OECD and beyond. 

The website follows a citizen-science approach, allowing every visitor to contribute to this research by reporting useful additional or new information. Should you have information on other AI tools used by tax administrations, contact us.​

State of use of AI by tax administrations

Since 2004, tax administrations are increasingly making use of machine-learning to perform their fiscal prerogatives.

AI has become an integral part of tax compliance risk-management as the primary technology to detect tax non-compliance.

Presently, every prerogative of a tax administration is underpinned by the use of AI systems and statistical models.

Tax administrations make use of AI to perform most of their missions:

• Taxpayer assistance: some AI systems, such as chatbots, voice bots, and 'friendly alerts' systems, have been integrated by tax administrations to provide voluntary assistance to taxpayers in need.
  These systems are meant to reduce the likelihood of errors when filing tax declarations in a non-coercive manner.
  

• Data collection: systems such as webscraping, image recognition, crawlers or ANPR, automatically collect taxpayer online data to match it with data already present in the data warehouses of tax administrations.
  These systems can recognise text, images, pictures and infer relevant information and signals of non-compliance for tax officials.
  

• Risk detection: tax administrations have integrated a range of systems to detect signals of non-compliance, suspicious transactions, unidentified taxpayers in real-time.
  Some of these systems can automatically block transactions, flag and freeze fraudulent accounts with little to no input from tax officials.
  

• Risk-management: these systems predict the risk of tax non-compliance associated with taxpayers and classify taxpayers according to their predicted risks.
  The predicted risk then helps tax officials in devising treatment strategies (deciding what action is appropriate for each taxpayer) and annual audit plans (deciding what taxpayer will be audited).
  

• Nudging: tax administrations also use AI to adapt the language of their communication to taxpayers, or adapt the design of their platforms to influence taxpayer behaviour without resorting to coercive means.
  In the context of tax compliance, nudging is for instance used to modify the language of default letters sent to taxpayers based on underlying profiling.
  

• Information management: to perform their prerogatives, tax administrations process billions of documents and trillions of data points every year. A number of AI systems assist tax officials in managing such large volumes of data.
  These systems help manage voluminous databases and find precise information within specific datasets, or help human resources to generate operational reports of actions performed.