Digital Tax Administration, AI and Emerging Compliance

Introduction and Context

Tax administration is moving from periodic, document-heavy compliance toward digitally connected, data-intensive and increasingly continuous supervision. E-filing, electronic invoicing, real-time reporting, digital payment records, automated risk assessment and artificial intelligence now influence how taxpayers report obligations and how tax authorities identify non-compliance.

For tax consultants, this development changes the nature of professional work. The consultant is no longer concerned only with preparing a return after the end of an accounting period. The professional must understand data flows, system controls, application programming interfaces (APIs), digital identity, transaction-level records, audit trails, information security and the governance of automated decisions.

Digitalisation can reduce compliance costs, improve accuracy and narrow the gap between tax declared and tax legally due. It can also create new risks. Incorrect master data may be reproduced across thousands of invoices; an unavailable platform may prevent timely filing; a poorly designed algorithm may produce discriminatory risk scores; and a cybersecurity incident may expose confidential taxpayer information. Effective digital tax compliance therefore requires a combination of tax knowledge, process design, technology governance and professional judgment.

This chapter examines the principal mechanisms of digital tax administration and develops a responsible framework for using automation and AI in tax compliance.

Learning Objectives

By the end of this chapter, the learner should be able to:

1. Explain the purposes and operating characteristics of e-filing, e-invoicing and real-time reporting.
2. Analyse how tax authorities match data from returns, invoices, payments, customs, payroll and third-party sources.
3. Design a continuous-controls approach to digital tax compliance.
4. Evaluate regulatory technology and automation used by taxpayers, advisers and tax authorities.
5. Apply principles of cybersecurity, privacy, explainability, human oversight and accountability to tax systems.
6. Assess emerging developments and recommend professionally responsible compliance practices.

1. From Periodic Filing to Digital Tax Administration

Traditional tax compliance is often represented as a cycle: transactions occur, accounts are closed, a return is prepared, and the authority later reviews the filing. Digital administration changes this sequence. Information may be created, validated, transmitted and compared at or near the time of the transaction.

Digital tax administration has four connected layers:

The essential principle is that compliance quality depends on the entire information chain, not merely on the final tax return. A correct return cannot reliably be produced from incomplete or corrupted source data.

Important terminology

• E-filing: Electronic submission of a tax return, claim, declaration or related document through an authorised digital channel.
• E-invoice: An invoice issued and exchanged in a structured electronic format that permits automated processing. A PDF sent by email may be electronic, but it is not necessarily a structured e-invoice.
• Real-time reporting: Transmission of specified transaction or tax data to a tax authority shortly after the relevant event, according to the applicable legal timetable.
• Digital tax administration: The use of digital infrastructure, data exchange, analytics and automated processes to administer tax obligations.
• Tax technology (TaxTech): Technology used by taxpayers and advisers to calculate, report, control or manage tax.
• RegTech: Technology that supports compliance with regulatory requirements, including monitoring, evidence collection and reporting.
• Tax gap: The difference between the tax legally due and the tax actually collected. Its measurement depends on the chosen estimation method and available data.
• Master data: Relatively stable reference data, such as taxpayer identification numbers, customer status, product codes, tax rates and jurisdictional classifications.
• Audit trail: A chronological record showing what was entered, changed, approved, transmitted or rejected, by whom and when.

2. E-Filing and Digital Identity

E-filing replaces paper or manually delivered returns with authenticated electronic submission. A mature e-filing system normally includes taxpayer registration, identity and access management, return preparation, validation, submission, acknowledgement, payment and record retention.

2.1 Benefits and control requirements

E-filing can provide:

• immediate confirmation of receipt;
• automatic arithmetic and completeness checks;
• pre-population from information already held by the authority;
• faster processing of refunds and assessments;
• standardised data suitable for analytics; and
• reduced printing, storage and administrative cost.

Its effectiveness depends on controls. A tax department should maintain a filing calendar, assign responsibility for each obligation, test digital credentials before deadlines and preserve submission acknowledgements. The acknowledgement should be retained with the submitted return, payment evidence and the version of supporting data used to prepare it.

Digital identity is particularly important. Authentication confirms that a user is authorised to access an account; authorisation determines what that user may do. These are different controls. A preparer may be authorised to draft a return but not to submit it, while a finance director may approve submission without editing underlying records.

A practical access model uses:

1. unique user accounts rather than shared credentials;
2. multi-factor authentication;
3. role-based access;
4. segregation of preparation, review and submission;
5. prompt removal of leavers' access; and
6. logs that cannot be casually altered.

2.2 E-filing failure management

A digital deadline remains a legal deadline unless the law or authority provides relief. Organisations should maintain documented procedures for platform outages, rejected files, invalid digital certificates and payment failures. Screenshots alone may not establish timely filing; reliable evidence may include system logs, transmission identifiers, authority acknowledgements, help-desk references and bank confirmations.

A sound contingency plan identifies:

• alternative submission channels, if legally available;
• escalation contacts;
• the person authorised to make a late-filing or outage claim;
• procedures for correcting duplicate submissions; and
• evidence-retention requirements.

3. E-Invoicing and Real-Time Reporting

E-invoicing is not simply the conversion of paper into a digital image. A structured invoice contains machine-readable fields such as supplier and customer identifiers, invoice number, date, taxable amount, tax rate, tax amount, currency, supply information and reference to related documents. The exact mandatory fields and transmission model are jurisdiction-specific.

Common operating models include:

1. Clearance model: The invoice or selected invoice data must be approved or validated by the authority before, or close to, issuance.
2. Central reporting model: The invoice is issued to the customer and relevant data is reported to the authority within a prescribed period.
3. Exchange-network model: Accredited service providers exchange structured invoices, sometimes with authority access to selected data.
4. Hybrid model: Different transaction classes or taxpayers use different combinations of clearance, reporting and periodic return requirements.

E-invoicing can improve input-tax verification, reduce fictitious invoicing and allow faster identification of missing trader patterns. It also creates operational risks. A wrong tax code, customer classification or place-of-supply decision can propagate automatically through the system.

3.1 Tax determination in an invoicing system

A simplified transaction calculation may be represented as:

Here, T is tax charged, B is the taxable base, and r is the applicable tax rate expressed as a decimal. The expression assumes that the transaction is taxable, the base is correctly determined, no special rounding or tax-inclusive pricing rule applies, and the rate is constant for the transaction. If a price is tax-inclusive, the extraction formula is different:

where G is the gross tax-inclusive amount. Both formulas are monetary calculations, so B, T and G must use the same currency unit. Rounding rules must be applied according to the relevant law or system specification, not assumed from ordinary arithmetic.

Worked example. A taxable sale has a net base of 12,000 currency units and a rate of 15%. Assuming no exemption, adjustment or special rounding rule:

T = 12,000 × 0.15 = 1,800

The gross amount is therefore 13,800 currency units. If 13,800 is instead a tax-inclusive amount, the taxable base would be 13,800 ÷ 1.15 = 12,000, and the tax would be 1,800. The same numerical result occurs here because the gross amount was constructed from the net amount; it should not be generalised to other figures.

3.2 Controls over e-invoices

Important controls include:

• sequential and unique invoice numbering;
• prevention of unauthorised deletion;
• controlled issuance of credit notes;
• validation of tax identification numbers;
• correct mapping of products and services to tax categories;
• reconciliation of invoices to the sales ledger and general ledger;
• monitoring of rejected or cancelled documents;
• testing of API transmissions; and
• preservation of original structured data and readable renderings.

A consultant should distinguish a technical transmission success from tax correctness. An authority may accept a file because it satisfies the schema while the underlying transaction remains incorrectly classified.

4. Tax Authority Data Matching and Risk Analysis

Tax authorities increasingly compare information from multiple sources. These may include tax returns, e-invoices, customs records, payroll submissions, banking information where legally available, property records, company registries, payment platforms and information exchanged internationally.

Data matching normally involves four stages:

1. Identification: Determine whether records relate to the same taxpayer, transaction or asset.
2. Standardisation: Convert dates, currencies, identifiers, codes and formats into comparable values.
3. Comparison: Test relationships, totals, timing and classifications.
4. Action: Request information, issue a notice, select an audit or close the case.

A simple reconciliation might compare reported sales with invoice data:

where S_r is sales reported in the return, S_i is sales represented in the relevant invoice dataset, and ∆ is the difference in currency units. The comparison is meaningful only when the periods, tax bases, inclusion rules, currencies and transaction populations are aligned. A difference is an investigation signal, not proof of evasion. Timing differences, exempt supplies, credit notes, imports, branch records and reporting thresholds may explain it.

Data quality can be assessed using dimensions such as accuracy, completeness, consistency, timeliness, validity and uniqueness. For example, duplicate taxpayer identifiers can cause false mismatches, while missing invoice dates can prevent correct period allocation.

4.1 Matching errors and professional response

False positives arise when compliant activity appears anomalous. Common causes include:

• common names or incorrectly entered identifiers;
• group-company transactions recorded in different periods;
• legitimate changes in business scale;
• currency conversion differences;
• credit notes issued after the original return; and
• classification differences between accounting and tax systems.

When responding to a data-matching query, the adviser should obtain the authority's stated population and period, reproduce the comparison independently, document reconciling items and submit evidence in a structured schedule. The response should not merely assert that the authority's data are wrong.

5. Continuous Controls and Continuous Compliance

Continuous compliance means that relevant controls operate throughout the reporting period rather than only during year-end preparation. It does not necessarily mean that every transaction is reviewed manually or that every tax issue is resolved instantly. It means that systems identify and address exceptions early enough to reduce accumulation of errors.

A continuous-control framework can be organised as follows:

5.1 Control design

Each material tax risk should have an owner, a defined control, a frequency, an evidence source, an exception threshold and an escalation route. A control description should state what is tested, who performs it, what system data are used and how the result is retained.

Examples include:

• daily comparison of invoice tax codes with approved product-tax mappings;
• weekly review of rejected e-invoices;
• monthly reconciliation of sales ledger, VAT return and general ledger;
• quarterly review of taxpayer master data; and
• automated alerts for unusual refunds, credit notes or jurisdictional combinations.

Automation should not conceal weak control design. A dashboard showing green indicators is not evidence that the underlying data are complete or that the thresholds are appropriate.

5.2 Exception management

A mature process classifies exceptions by severity. A failed transmission affecting one low-value invoice may be operationally significant but different from a systematic misclassification affecting an entire product line. Each exception should have a status such as open, investigated, corrected, accepted with rationale or escalated.

Useful management measures include the number of exceptions, age of unresolved exceptions, recurrence rate, percentage resolved within target time and value exposed. These are management indicators rather than universal legal standards; their definitions must be documented so that trends are comparable.

6. Automation, AI and Regulatory Technology

Automation follows explicit rules. For example, a system may reject a return when a mandatory field is blank. Artificial intelligence generally refers to computational systems that perform tasks associated with functions such as prediction, classification, pattern recognition, language processing or generation. Machine learning systems infer patterns from data rather than relying exclusively on hand-written rules.

In tax compliance, possible uses include:

• extracting data from invoices and contracts;
• classifying transactions for review;
• identifying anomalous refund or input-tax claims;
• forecasting filing workload;
• matching entities across datasets;
• drafting explanations or correspondence; and
• prioritising human review.

These uses differ in risk. A tool that extracts invoice dates for human confirmation is generally less consequential than a model that recommends denial of a refund or selects a taxpayer for intrusive audit.

6.1 A risk-based AI governance model

A responsible lifecycle contains seven stages:

1. Purpose definition: Specify the lawful business or administrative purpose and prohibited uses.
2. Data assessment: Examine provenance, quality, representativeness, retention and access rights.
3. Model or rule design: Document inputs, transformations, assumptions and output meaning.
4. Validation: Test accuracy, false positives, false negatives, stability and relevant subgroups.
5. Deployment controls: Limit access, record versions, establish thresholds and preserve logs.
6. Human review: Ensure that consequential decisions can be examined and challenged.
7. Monitoring and retirement: Detect drift, investigate incidents and withdraw systems that no longer perform acceptably.

6.2 Explainability and human oversight

Explainability does not require every model to be mathematically simple, but it does require an intelligible account of the factors that materially influenced an output and the limits of that output. A taxpayer or reviewer should be able to understand whether a result is a lead, a prediction, a legal conclusion or an automated decision.

Human oversight must be substantive. A reviewer who merely approves an algorithmic recommendation without access to relevant evidence is not exercising meaningful judgment. Reviewers require authority, time, competence and a route to override or correct the system.

AI-generated tax advice also requires special caution. Generative systems may produce plausible but incorrect interpretations, outdated rates or invented authorities. Professional users should verify primary legal sources, identify the date and jurisdiction of the law, preserve prompts and outputs where material, and never treat generated text as a substitute for professional responsibility.

7. Cybersecurity, Privacy and Information Governance

Tax data are highly sensitive. They may reveal income, payroll, customers, suppliers, assets, health-related claims or business strategy. Digital compliance therefore requires confidentiality, integrity and availability.

7.1 Core safeguards

A proportionate security programme should include:

• data classification and least-privilege access;
• multi-factor authentication;
• encryption in transit and at rest where appropriate;
• secure API credentials and key rotation;
• network and endpoint protection;
• vulnerability management and patching;
• tested backups and recovery procedures;
• supplier due diligence;
• phishing and social-engineering training;
• immutable or protected audit logs; and
• incident detection, reporting and response.

Security controls should be tested, not merely documented. A recovery plan is credible only if restoration has been exercised and recovery times are understood.

7.2 Privacy principles

Privacy regulation differs by jurisdiction, but widely recognised principles include lawfulness, purpose limitation, data minimisation, accuracy, storage limitation, security, transparency and accountability. Tax compliance may provide a legal basis for processing, but it does not justify unlimited collection or uncontrolled secondary use.

Before introducing an AI or analytics project, the organisation should ask:

• What precise purpose requires the data?
• Is each data field necessary?
• What authority permits collection and disclosure?
• How long will information be retained?
• Who can access it, including vendors?
• Can individuals challenge inaccurate data or consequential outcomes?
• What happens if the system is wrong?

Cross-border cloud hosting and international data transfers require specific legal and contractual analysis. Confidentiality obligations may also arise from professional conduct rules, engagement letters and tax secrecy provisions.

8. Implementation Framework for Tax Consultants

A tax consultant supporting digital compliance should begin with a process and risk inventory rather than immediately selecting software.

Step 1: Map obligations and data flows

List returns, payments, invoices, certificates, reconciliations, registrations and evidence requirements. Map the path from source transaction to accounting entry, tax calculation, submission and archive.

Step 2: Identify failure points

Consider incorrect master data, missing transactions, duplicate records, late approvals, interface failures, unauthorised changes, incorrect period allocation and weak evidence retention.

Step 3: Classify controls

Separate preventive, detective, corrective and monitoring controls. Identify which are automated and which require human judgment.

Step 4: Establish ownership and segregation

Assign process owners, reviewers, approvers and technical administrators. Avoid allowing one person to create a vendor, alter tax settings and approve payment without independent review.

Step 5: Test before deployment

Use representative transactions, including exemptions, credit notes, foreign currency, imports, partial recovery and unusual dates. Test both successful and rejected messages.

Step 6: Document evidence

Retain configuration versions, test scripts, approvals, exception reports, reconciliations, submission acknowledgements and change records. Evidence should show not only what the system produced but why it was considered reliable.

Step 7: Monitor and improve

Review error trends, authority feedback, security incidents, legislative changes and model performance. Update controls when products, jurisdictions or reporting rules change.

9. Illustrative Case: Northstar Components Ltd.

Background

Northstar Components Ltd. is a medium-sized manufacturer selling domestically and across two neighbouring jurisdictions. It introduces structured e-invoicing and monthly indirect-tax reporting. Its enterprise system contains product codes, customer status, delivery locations and tax rules.

Problem

After implementation, the tax authority reports that Northstar's invoice data indicate taxable domestic sales exceeding sales reported in its return by 4.2 million currency units. The authority requests an explanation and warns that the difference may indicate under-reporting.

Analysis

The consultant does not immediately accept or reject the authority's figure. The investigation proceeds in stages:

1. The periods are aligned. The authority's invoice dataset covers invoices issued during the month, while Northstar's return is based on tax point rules that include certain invoices from the preceding period.
2. Credit notes are identified. Several credit notes were issued in the following month but related to the earlier period.
3. Export invoices are separated from domestic taxable sales. The system had correctly marked them as exports, but the authority's initial comparison included them in gross invoice value.
4. A product mapping error is found. A newly acquired product line was assigned a domestic tax code, although some sales qualified for a different treatment subject to documentary conditions.
5. Transmission logs show that two batches were rejected and successfully retransmitted. The accounting ledger contains one original and one corrected record, but the reporting extract counted both.

Approach and solution

Northstar prepares a bridge schedule from gross invoice data to reported taxable sales. Each reconciling item is supported by invoice numbers, credit notes, export evidence, tax-point analysis and transmission logs. The company corrects the duplicate extract, reviews the product mapping, and submits an amended return for the genuinely misclassified transactions. It also introduces a daily rejected-file report and a monthly three-way reconciliation between the invoice register, general ledger and tax return.

The AI component is limited to prioritising unusual combinations of product, destination and tax code for human review. It does not automatically change tax treatment. The model's inputs, version, review outcomes and override reasons are recorded.

Outcome

The authority accepts most of the reconciliation and opens a narrower review of the product-line classification. Northstar avoids treating the original data-matching signal as a final conclusion, while still correcting a real error. The new controls detect similar mapping problems before the next filing.

Learning points

• A data mismatch is a starting point for investigation, not conclusive evidence of non-compliance.
• Period, population and classification definitions must be aligned before comparison.
• Digital systems can reproduce both accurate and inaccurate rules at scale.
• Transmission logs and versioned evidence are essential in explaining digital records.
• AI is safer when used to prioritise human review than when allowed to make unreviewed legal determinations.

10. Challenges, Limitations and Regulatory Developments

Interoperability

Different authorities and trading partners may require different schemas, identifiers, authentication methods and reporting timetables. Interoperability failures can create duplicate work and inconsistent records. Organisations should use controlled mappings and avoid uncontrolled spreadsheet conversions.

Legacy systems and technical debt

Older accounting systems may not store the fields required for real-time reporting. A new interface may therefore transfer incomplete data while appearing technically successful. Data lineage and reconciliation are necessary whenever systems are integrated.

Algorithmic bias and opacity

Historical enforcement data may reflect past institutional priorities or unequal reporting behaviour. A model trained on such data can reproduce those patterns. Accuracy alone is insufficient; fairness, proportionality, legal authority and contestability must also be considered.

Model drift

Business models, tax law, fraud patterns and taxpayer behaviour change. A model that performed well during one period may deteriorate later. Performance monitoring should define when retraining, recalibration or withdrawal is required.

Over-reliance on automation

Automation can encourage staff to trust system outputs without understanding assumptions. Professional skepticism remains necessary, particularly where the system makes legal classifications, estimates liability or evaluates unusual transactions.

Regulatory evolution

Future compliance is likely to involve wider structured reporting, digital identity, platform reporting, automated exchange of information, electronic audit files, machine-readable legislation and more frequent authority feedback. The exact legal requirements will differ by jurisdiction. Tax professionals should monitor legislation, technical specifications, official guidance, enforcement practice and professional standards rather than relying on generic technology claims.

Best-Practice Checklist

A digitally mature tax function should be able to answer yes to the following questions:

• Are all tax obligations mapped to named owners and deadlines?
• Are source data, transformations and submissions traceable?
• Are tax codes and taxpayer master data subject to controlled change?
• Are e-invoice rejections, duplicates and cancellations monitored?
• Are returns reconciled to underlying ledgers and relevant third-party data?
• Are system access, credentials and segregation of duties reviewed?
• Are AI purposes, data sources, versions and limitations documented?
• Can a human explain, challenge and correct a material automated output?
• Are privacy, retention, vendor and cross-border data issues assessed?
• Have business continuity and cyber-incident procedures been tested?

Chapter Summary

Digital tax administration connects transaction systems, structured tax documents, electronic filing, authority data platforms and analytical tools. E-filing improves speed and standardisation but requires identity, access, submission and evidence controls. E-invoicing and real-time reporting provide more timely information, yet make accurate master data, tax coding, interfaces and exception management essential.

Tax authority data matching can identify inconsistencies across returns, invoices, payroll, customs and other lawful data sources. A mismatch is an investigative signal rather than automatic proof of wrongdoing; professional analysis must align periods, populations, tax rules and data definitions. Continuous compliance uses preventive, detective, corrective and monitoring controls throughout the reporting cycle.

Automation and AI can support extraction, classification, anomaly detection and workflow prioritisation. Their use must be governed through clear purpose, reliable data, validation, version control, human oversight, explainability, monitoring and accountability. Cybersecurity and privacy are central because tax data are sensitive and digital systems create concentrated operational risk.

The future tax professional will need integrated competence in tax law, accounting data, systems controls, cybersecurity, privacy and responsible technology governance. Digital tools can improve compliance, but they do not transfer legal or professional responsibility from the taxpayer or adviser to the software.

Key Terms

E-filing: Electronic submission of a tax return or related document through an authorised channel.

E-invoice: A structured electronic invoice designed for automated exchange and processing.

Real-time reporting: Transmission of specified transaction or tax information shortly after an event.

Digital tax administration: Technology-enabled administration using digital records, data exchange and analytics.

TaxTech: Technology used to calculate, manage, report or control tax obligations.

RegTech: Technology supporting regulatory compliance, monitoring and reporting.

Master data: Reference data such as taxpayer identifiers, product codes and tax classifications.

Audit trail: Chronological evidence of system actions, changes, approvals and transmissions.

Data matching: Comparison of records from different sources to identify relationships or discrepancies.

Continuous compliance: Ongoing operation and monitoring of tax controls during the reporting period.

Exception management: The process of identifying, investigating, correcting and escalating control failures.

Artificial intelligence: Computational capability associated with tasks such as prediction, classification, language processing or pattern recognition.

Machine learning: A method in which a system infers patterns from data to produce outputs.

Explainability: The ability to communicate how an automated output was produced and what its limitations are.

Human oversight: Meaningful human review, challenge and intervention in automated processes.

Data lineage: The traceable path of data from source through transformations to final output.

Least privilege: Granting users only the access necessary for their authorised duties.

Model drift: Decline in model relevance or performance as data, behaviour or conditions change.

Privacy by design: Incorporating privacy protections into systems and processes from the beginning.

Tax gap: The difference between tax legally due and tax actually collected.

Review Questions

Short Answer Questions

1. Distinguish between an electronic invoice and a structured e-invoice.
2. Why is a data-matching discrepancy not automatically proof of tax evasion?
3. State two controls that support secure e-filing.
4. What is the difference between preventive and detective tax controls?
5. Why is human oversight necessary when AI is used in tax compliance?

Long Answer Questions

1. Explain how e-filing, e-invoicing and real-time reporting collectively change the tax compliance cycle.
2. Discuss the principal data-quality risks in tax authority matching and propose professional responses.
3. Develop a continuous-controls framework for a business subject to electronic indirect-tax reporting.
4. Evaluate the benefits and limitations of using AI to identify anomalous tax transactions.
5. Explain how cybersecurity, privacy and professional accountability should influence a digital tax implementation.

Analytical / Application Questions

1. A company’s invoice extract exceeds reported taxable sales by 8%. Design an investigation that distinguishes timing, export, credit-note, duplicate-record and genuine-understatement explanations.
2. A tax authority proposes using a machine-learning model to select taxpayers for audit. Identify the governance, fairness, explainability, data-quality and human-review controls that should be required before deployment.
3. Advise a business whose e-invoicing system technically transmits documents successfully but frequently assigns incorrect tax codes. Recommend immediate corrective actions and longer-term control improvements.

Multiple Choice Questions with Answers

Question 1

Which feature most clearly distinguishes a structured e-invoice from a PDF invoice?

A. It is sent by email.

B. It contains machine-readable fields for automated processing.

C. It includes a company logo.

D. It can be printed.

Correct Answer: B. It contains machine-readable fields for automated processing.

Question 2

What is the primary purpose of multi-factor authentication in e-filing?

A. To calculate tax automatically

B. To replace segregation of duties

C. To strengthen verification of user identity

D. To reconcile invoices with the ledger

Correct Answer: C. To strengthen verification of user identity

Question 3

A tax authority finds a difference between return sales and invoice data. What is the most appropriate initial conclusion?

A. The taxpayer has committed evasion.

B. The data are unusable in every case.

C. The difference is an investigation signal requiring reconciliation.

D. The return must automatically be cancelled.

Correct Answer: C. The difference is an investigation signal requiring reconciliation.

Question 4

Which is a preventive control?

A. A post-period exception report

B. Blocking an invalid tax code before posting

C. Investigating a tax authority query

D. Reviewing last year’s audit findings

Correct Answer: B. Blocking an invalid tax code before posting

Question 5

Using the formula T = B × r, what is the tax on a base of 20,000 at a rate of 10%, assuming no adjustments?

A. 200

B. 1,000

C. 2,000

D. 20,100

Correct Answer: C. 2,000

Question 6

Which AI use generally presents the highest need for human oversight?

A. Suggesting invoice fields for confirmation

B. Sorting documents by date

C. Automatically denying a taxpayer’s refund claim

D. Detecting duplicate file names

Correct Answer: C. Automatically denying a taxpayer’s refund claim

Question 7

What does data lineage describe?

A. The legal rate of tax

B. The path of data from source through transformations to output

C. The number of users in a system

D. The age of a taxpayer account

Correct Answer: B. The path of data from source through transformations to output

Question 8

Which principle limits collection of unnecessary personal information?

A. Data minimisation

B. Unlimited retention

C. Open access

D. Automatic profiling

Correct Answer: A. Data minimisation

Question 9

What is model drift?

A. A cyberattack on a tax portal

B. A decline in model relevance as conditions change

C. A duplicate invoice number

D. A type of tax exemption

Correct Answer: B. A decline in model relevance as conditions change

Question 10

Which practice best supports reliable digital tax compliance?

A. Allowing shared administrator passwords

B. Relying only on a green dashboard indicator

C. Reconciling source records, ledgers and submitted returns

D. Deleting rejected transmission records

Correct Answer: C. Reconciling source records, ledgers and submitted returns