tA ble of content S
I.1.2.1.
I.1.2.2.2. Guidelines of the Tax Agency’s Strategic Plan 2020–2023 ............................................ 76
I.1.2.2.3. Guidelines of the Tax Agency’s Strategic Plan 2024–2027 ............................................ 100
I.2. An approach to mass data processing systems and the potential use of artificial intelligence .....................................
I.2.1. Mass data processing systems ...........................
I.2.1.1. The Zújar ecosystem: the great corporate analytics engine .............................................
I.2.1.2. Specialised and complementary analytical environments
I.2.1.3. Capabilities for unstructured information, advanced relational analysis and forensic computing
I.2.1.4. Massive data sources and essential integration tools
I.2.1.5. The underlying technological infrastructure for big data processing ..........................................
I.2.2. Detailed classification of systems by likelihood of using artificial intelligence techniques ..................................... 139
I.2.2.1. Confirmed artificial intelligence techniques or explicit or acknowledged use ...................................
I.2.2.2. Artificial intelligence that is highly probable or whose functionality strongly suggests its use ....................
I.2.2.3. Use of artificial intelligence as a possible or potential ‘ ’ 159
I.2.3. The state of artificial intelligence: the gap between evidence and official discourse
I.3. The Artificial Intelligence Strategy of the State Tax Administration Agency
I.3.1. Structure and content of the Artificial Intelligence Strategy and the ethical commitment
I.3.1.1. The principle of proactive responsibility
I.3.1.2. A human-centric approach
I.3.1.3. The security and governance of artificial intelligence ... 173
I.3.2. Selective transparency and the imbalance between assistance and control .................................................
I.3.3. The practical implementation of ethical principles and fundamental safeguards ..............................................
I.3.4. Strategy as (show-off) positioning vs. (effective) operational governance ..................................................
I.3.5. A critical approach ................................... 182
c h A pter ii .
II.1. The restrictive interpretation of the concept of an artificial intelligence system: a headlong rush in the actions of the State Tax Administration Agency? ................................................ 185
II.2. The seven defining elements of artificial intelligence systems ....... 194
II.2.1. They are machine-based systems ........................ 194
II.2.2. They are designed to operate with varying degrees of autonomy . .
195
II.2.3. They may demonstrate adaptability following deployment..... 199
II.2.4. They pursue explicit and implicit objectives ............... 201
II.2.5. They infer and generate outputs ........................ 203
II.2.6. Predictions, content, recommendations and/or decisions ...... 206
II.2.7. Outputs may influence physical or virtual environments ...... 209
II.2.8. Systems that fall outside the concept of artificial intelligence .. 210
II.3. Extending the definition of artificial intelligence systems to the big data processing applications used by the State Tax Administration Agency . 212
II.3.1. Data mining system for the Register of Intra-Community Operators . .
II.3.2. Value Added Tax refund system for non-EU residents ........ 216
II.3.3. MIDAS Project ...................................... 218
II.3.4. Virtual assistants and conversational technologies (AVIVA, Income Tax Assistant, Census Assistant, TEAC/DGT Assistants, Electronic Office Assistant, IVR) ......................................
II.3.5. NIDEL ............................................
II.3.6. Specific personal income tax predictive models (detection of non-filers) ...............................................
II.3.7. DOBLING customs system ............................
II.3.8. Corpus preparation for large language models (generative AI)
II.3.9. Assistance/error reduction in personal income tax and income tax nudges
II.3.10. Buscón (corporate search and indexing engine)
II.3.11. Advanced forensic analytics tools (Cellebrite Pathfinder Enterprise)
II.3.12. Customs risk analysis algorithm
II.3.13. HERMES (risk management and analysis system)
II.3.14. Crypto-asset analysis software
II.3.15. HERACLES Project
II.3.16. TESEO (graphical link analysis tool)
II.3.17. RIFA (Open-Source Information Retrieval)
II.3.18. System for detecting ‘false non-residents’
II.3.19. TNA (Transaction National Analysis)
II.3.20. Economic capacity data analysis service
II.3.21. Tax Intelligence System (ALPHA)
II.3.22. Unified Knowledge Base (BUC) – Legal Intelligence
II.3.23. Comprehensive Claims Management System (PLATEA)
II.3. Reflection
c h A pter iii . t he cl ASSific Ation of A rtifici A l intelligence S yS tem S ..........
III.1. High-risk artificial intelligence systems at the State Tax Administration Agency? ................................................ 255
III.1.1. The concept of high risk ............................. 255
III.1.2. Exclusion criteria for high-risk systems .................. 261
III.1.2.1. The absence of substantial influence .............. 262
III.1.2.2. Profiling: automatic inclusion criterion ............. 263
III.1.3. High-risk systems for access to and use of essential services and benefits ............................................ 264
III.1.3.1. Context and regulatory developments .............. 265
III.1.3.2. The restrictive interpretation .................... 269
III.1.4. High-risk systems used to ensure compliance with the law ... 271
III.1.4.1. Context and regulatory developments 271
III.1.4.2. Inferences of criminal conduct obtained through administrative channels by artificial intelligence systems: a high risk? 277
III.1.4.3. The restrictive interpretation: could the concept of high risk linked to criminal offences lead to disparities in the level of protection across different EU countries? ................ 280
III.2. Artificial intelligence systems prohibited in the tax sphere ......... 281
III.2.1. Subliminal, deliberate or misleading manipulation ......... 282
III.2.1.1. Concept and requirements....................... 282
III.2.1.2. Incentives in the administrative criterion (non-modifiable) on Renta Web .................................. 285
III.2.1.3. Information tools with biased interpretations ....... 286
III.2.1.4. Tax actions based on risk profiles ............... 287
III.2.1.5. The application of the prohibition on manipulation and harmful deception 289
III.2.2. Exploitation of vulnerabilities 290
III.2.2.1. Concept and requirements 290
III.2.2.2. The governance gap and the lack of protection for vulnerable individuals 292
III.2.3. Generalised social scoring 294
III.2.3.1. Concept and requirements ...................... 294
III.2.3.2. Legitimate assessment for specific purposes vs. prohibited social scoring ................................... 297
III.2.3.3. Social scoring in the context of the tax authorities ... 298
III.2.4. Risk assessment to predict crime based solely on profiling or personality .............................................. 300
III.2.4.1. Concept and requirements ...................... 300
III.2.4.2. Applicability to risk analysis systems in the tax context 301
III.2.5. Creation/expansion of facial databases through non-selective scraping, emotion inference and real-time remote biometric identification 302
III.2.6. The limited impact of the AI Act prohibited practices on the STA
III.3. Low-risk or minimal-risk systems ............................
III.3.1. Limited-risk systems: the emphasis on transparency .........
III.3.2. Systems of minimal or zero risk: the absence of specific AI Act obligations ...........................................
III.4. Classification of the systems analysed and consequences ..........
III.4.1. Classification of the systems analysed ...................
III.4.2. Consequences of the classification ......................
III.4.3. Governance and supervision of artificial intelligence in Spain: the important role of the AESIA in the classification of artificial intelligence systems and its challenges regarding independence ........
IV.1. The deployment of artificial intelligence in the application of the tax system: the duty to contribute to the funding of public expenditure and the impact on legally protected interests arising from potential conflicts ................................................
IV.1.2.
IV.1.3.
IV.1.4.
IV.1.5. Effective judicial protection and the rights of defence
IV.1.6.
IV.1.7.
IV.2. The inadequacy of control mechanisms. The fragmented regulation of artificial intelligence systems through domestic law and its implications for their management by the tax authorities
IV.2.1. General tax and administrative regulations
IV.2.1.1.
IV.2.1.1.1.
IV.2.1.1.2. The main limitations on ensuring transparency and
IV.2.1.1.3. The inadequacy of the specific tax regulatory framework in relation to artificial intelligence systems
IV.2.1.2. Automated administrative action in the general administrative sphere .....................................
IV.2.1.3. The Supreme Court’s ruling on the BOSCO case and its implications for the tax sphere ..........................
IV.2.1.4. The Elsbury doctrine and the rejection of opacity as a deterrent strategy
IV.2.2. The implications of data protection legislation ............. 397
IV.2.2.1. Their inadequacy in the face of the emergence of artificial intelligence systems: key elements ...................
IV.2.2.2. Two further key issues .........................
IV.2.2.2.1. The absence of impact assessments as a shortcoming in ex ante control ............................
IV.2.2.2.2. The inadequate quality of the law ............
IV.2.3. The Comprehensive Law on Equal Treatment and Non-Discrimination
IV.2.4. Soft law
IV.2.5. The manifest inadequacy of the current regulatory framework
IV.3. The defence of taxpayers against the use of artificial intelligence systems by tax authorities: where are the legal safeguards? Possible avenues for challenge
IV.3.1. Key challenges for the defence
IV.3.2. The barrier of algorithmic opacity
IV.3.3. Insufficiency of formal reasoning
IV.3.4. Difficulty in accessing information and evidence
IV.3.5. De facto shift in the burden of proof
Potential ineffectiveness of essential procedures
V.1. Regarding the evolution of artificial intelligence systems and mass data processing at the State Tax Administration Agency (Chapter I)
Table 4. Critical analysis of the eiaat and the ethical commitment (rhetoric vs.
Table 5. Summary. Classification of sta systems as ais under the ai act (art. 3.1)
XII. Computer Forensics and OSINT ..............................
A nnex iii .
glo SSA ry of term S rel Ated to A rtifici A l intelligence S yS tem S ..
Bernardo D. Olivares Olivares
