AI adoption in tax administration
These answers come from the OECD's Inventory of Tax Technology Initiatives (ITTI), in which each administration describes its own technology use. This page reproduces those answers and labels them as self-reported โ it does not independently verify what any authority actually does with AI. A "No" means the administration reported no such use at the time it answered the survey, not that such use is impossible or ruled out.
Administrations reporting they use artificial intelligence
| Country | Reported answer |
|---|---|
| ๐ฆ๐ฑ Albania | Yes |
| ๐ฆ๐ฒ Armenia | Yes |
| ๐ฆ๐บ Australia | Yes |
| ๐ฆ๐น Austria | Yes |
| ๐ฆ๐ฟ Azerbaijan | Yes |
| ๐ง๐ท Brazil | Yes |
| ๐ง๐ฌ Bulgaria | Yes |
| ๐จ๐ฆ Canada | Yes |
| ๐จ๐ณ China | Yes |
| ๐จ๐ด Colombia | Yes |
| ๐จ๐ท Costa Rica | Yes |
| ๐ญ๐ท Croatia | Yes |
| ๐จ๐บ Cuba | Yes |
| ๐ฉ๐ฐ Denmark | Yes |
| ๐ช๐จ Ecuador | Yes |
| ๐ช๐ช Estonia | Yes |
| ๐ซ๐ฎ Finland | Yes |
| ๐ซ๐ท France | Yes |
| ๐ฉ๐ช Germany | Yes |
| ๐ฌ๐ท Greece | Yes |
| ๐ญ๐ฐ Hong Kong SAR, China | Yes |
| ๐ญ๐บ Hungary | Yes |
| ๐ฎ๐ฉ Indonesia | Yes |
| ๐ฎ๐ช Ireland | Yes |
| ๐ฎ๐ฑ Israel | Yes |
| ๐ฎ๐น Italy | Yes |
| ๐ฏ๐ต Japan | Yes |
| ๐ฐ๐ท Korea, Rep. | Yes |
| ๐ฑ๐ป Latvia | Yes |
| ๐ฑ๐ธ Lesotho | Yes |
| ๐ฑ๐น Lithuania | Yes |
| ๐ฒ๐พ Malaysia | Yes |
| ๐ฒ๐ฑ Mali | Yes |
| ๐ฒ๐ณ Mongolia | Yes |
| ๐ณ๐ฟ New Zealand | Yes |
| ๐ณ๐ด Norway | Yes |
| ๐ต๐ญ Philippines | Yes |
| ๐ต๐ฑ Poland | Yes |
| ๐ต๐น Portugal | Yes |
| ๐ธ๐ฆ Saudi Arabia | Yes |
| ๐ธ๐ฌ Singapore | Yes |
| ๐ธ๐ฎ Slovenia | Yes |
| ๐ฟ๐ฆ South Africa | Yes |
| ๐ช๐ธ Spain | Yes |
| ๐ธ๐ช Sweden | Yes |
| ๐น๐ผ Taiwan | Yes |
| ๐น๐ฏ Tajikistan | Yes |
| ๐น๐ญ Thailand | Yes |
| ๐น๐ท Turkiye | Yes |
| ๐ฆ๐ช United Arab Emirates | Yes |
| ๐ฌ๐ง United Kingdom | Yes |
| ๐บ๐ธ United States | Yes |
| ๐ฟ๐ฒ Zambia | Yes |
AI used for detection of tax evasion and fraud
| Country | Reported answer |
|---|---|
| ๐ฆ๐ฑ Albania | Yes |
| ๐ฆ๐ฒ Armenia | Yes |
| ๐ฆ๐บ Australia | Yes |
| ๐ฆ๐น Austria | Yes |
| ๐ฆ๐ฟ Azerbaijan | Yes |
| ๐ง๐ท Brazil | Yes |
| ๐จ๐ฆ Canada | Yes |
| ๐จ๐ด Colombia | Yes |
| ๐จ๐ท Costa Rica | Yes |
| ๐จ๐บ Cuba | Yes |
| ๐ฉ๐ฐ Denmark | Yes |
| ๐ซ๐ฎ Finland | Yes |
| ๐ซ๐ท France | Yes |
| ๐ฉ๐ช Germany | Yes |
| ๐ฌ๐ท Greece | Yes |
| ๐ญ๐บ Hungary | Yes |
| ๐ฎ๐ฉ Indonesia | Yes |
| ๐ฎ๐ฑ Israel | Yes |
| ๐ฎ๐น Italy | Yes |
| ๐ฏ๐ต Japan | Yes |
| ๐ฑ๐น Lithuania | Yes |
| ๐ฒ๐พ Malaysia | Yes |
| ๐ฒ๐ฑ Mali | Yes |
| ๐ฒ๐ณ Mongolia | Yes |
| ๐ณ๐ฟ New Zealand | Yes |
| ๐ณ๐ด Norway | Yes |
| ๐ต๐ญ Philippines | Yes |
| ๐ต๐ฑ Poland | Yes |
| ๐ธ๐ฆ Saudi Arabia | Yes |
| ๐ธ๐ฌ Singapore | Yes |
| ๐ธ๐ฎ Slovenia | Yes |
| ๐ฟ๐ฆ South Africa | Yes |
| ๐ช๐ธ Spain | Yes |
| ๐ธ๐ช Sweden | Yes |
| ๐น๐ผ Taiwan | Yes |
| ๐น๐ญ Thailand | Yes |
| ๐ฌ๐ง United Kingdom | Yes |
| ๐บ๐ธ United States | Yes |
| ๐ฟ๐ฒ Zambia | Yes |
AI involved in making final administrative decisions
This is a different and more consequential thing than AI assisting a human reviewer: it means the administration itself reported that AI plays a role in producing the decision that is final, not a recommendation a person then signs off on. The list below reports exactly what each administration told the OECD โ it is not an assessment of due-process safeguards, appeal rights, or how much human review actually sits around that AI step.
| Country | Reported answer |
|---|---|
| ๐ฆ๐ฑ Albania | Yes |
| ๐ฆ๐ฒ Armenia | Yes |
| ๐ฆ๐ฟ Azerbaijan | Yes |
| ๐ฟ๐ฒ Zambia | Yes |