Table 4.
Mapping the main features of Brazilian AI-ACTs used to monitor governmental revenue, the management of public funds, and citizens’ complaints and denunciations
Specific areas | Focal point (Number of initiatives) | Name of the AI-ACT/Developer | Approach | Users | Promote impact via | Main weaknesses observed |
---|---|---|---|---|---|---|
Public revenue (e.g., tax evasion or avoidance) | Customs clearance issues (3) | Sisam, Aniita, Ajna/Revenue Service | Top-down | Revenue Service Inspectors | Automatizing and standardizing administrative tasks; assisting financial investigations | Not fully trusted by users; alert overflow; existing false positives; slow adaptability to new types of wrongdoing; limited auditability; high risk of bias and unfairness |
Incompatible financial transactions (1) | ContÁgil/Revenue Service | Top-down | Revenue Service Inspectors | Automatizing and standardizing administrative tasks; assisting financial investigations | Not fully trusted by users; alert overflow; existing false positives; slow adaptability to new types of wrongdoing; limited auditability; high risk of bias and unfairness | |
Management of public funds and reports on finances | Fiscal irresponsibility spotted and sanctioned by the Court of Accounts (1) | Zello Chatbot/TCU | Top-Down | Any citizen with access to Twitter and/or WhatsApp | Promoting downward transparency; automatizing and standardizing administrative tasks | Limited data; limited auditability; issues to engage the ordinary citizen; lack of use |
General | Complaints and denunciations of any type of irregularity (2) | FARO/CGU | Top-down | Auditors from the CGU | Automatizing and standardizing administrative tasks; promoting upward transparency | Lack of use due to the high number of false positives; slow adaptability to new types of wrongdoing; limited auditability; risk of unfairness |
Cida Chatbot/CGU | Top-down | Any citizen with access to Facebook and Telegram | Automatizing and standardizing administrative tasks; promoting upward transparency | Limited data; issues to engage the ordinary citizen; lack of use |