The article asks how the intelligent tools used by tax authorities should be judged, and who answers when something goes wrong. Public tax intelligence tools are treated as the face of smart taxation, serving tax authorities, taxpayers and third-party intermediaries. The author lists three weak spots: standards for collecting tax-related information are unclear and the authority behind them thin; information management lacks transparency and publicness; and information protection suffers from coarse compliance auditing and weak constraints on algorithmic risk. A four-step benchmark is proposed, moving from tax simplification through tax transparency and tax legality to tax choice, used as the yardstick in pre-release testing. Once released, liability should be allocated differently for narrow and general AI settings, in related but distinct ways.
News Detail
Evaluation Benchmarks and Liability Allocation for Public Tax Intelligence Tools
An article in the July 2024 issue of Taxation Research looks at how public tax intelligence tools should be evaluated and how liability should be allocated. It treats these tools as the front end of smart taxation, serving tax authorities, taxpayers and third-party tax intermediaries. Three problem areas are identified: unclear standards and weak authority for collecting tax-related information; limited transparency and insufficient publicness in information management; and inadequate compliance auditing plus weak rules on algorithmic risk in information protection. The author proposes a four-step benchmark proceeding from tax simplification to tax transparency, tax legality and tax choice, used as pre-release testing criteria. After release, liability should be designed separately for narrow and general AI contexts, in ways that are related yet distinct.
2026-11-04