Israeli Tax Authority Reportedly Used an Opaque Automated System to Issue a Fine, Declining to Explain or Disclose the Underlying Calculation

An Israeli farmer, Moshe Har Shemesh, reportedly received a fine generated by a Tax Authority software system whose calculation officials were allegedly unable to explain. When the farmer reportedly sought access to the program or its source code to understand the basis for the amount, the authority allegedly refused, citing security concerns and the difficulty of extracting the embedded guidelines. The dispute reportedly later moved into legal proceedings focused on whether code and automated decision rules constitute information subject to public disclosure.

Risk classification

• Primary risk domain: 7 AI system safety, failures, & limitations
• Primary risk subdomain: 7.4 Lack of transparency or interpretability

The core of the dispute is the Tax Authority's inability to explain the automated fine and its refusal to disclose the software's source code, preventing public oversight.

Additional risk subdomains

• 5.2 Loss of human agency and autonomy: The Tax Authority delegated administrative discretion to automated software without public discussion or human review of the specific calculations.
• 7.3 Lack of capability or robustness: The software exhibited a biased calculation error in VAT recalculations, leading to a court-ordered compensation of 2 million shekels.

Causal factors

• Entity: AI
• Intent: Unintentional
• Timing: Post-deployment

The incident was caused by the automated tax software's opaque calculations and biased VAT adjustments occurring after its deployment.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The system is used by a public authority to make decisions on tax fines and financial sanctions, which has significant implications for fundamental rights, public interest, and civil oversight.

AI system and alleged parties

• AI system: Tax Authority computer software
• AI purpose: Financial Processing
• Behaviour type: Autonomous
• Alleged developer: Israeli Tax Authority
• Alleged deployer: Israeli Tax Authority
• Alleged harmed parties: Moshe Har Shemesh, Israeli people having tax fines

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Minor. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Minor, indirect Minor
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Minor, indirect Negligible
• Civil rights: direct Minor, indirect Minor
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Financial

Reported: The report explicitly describes financial losses, including an unexplained fine for Har Shemesh farm and a 2 million shekel compensation payment ordered by a court in a related class action lawsuit.

Directly caused: The court ordered Israel to pay compensation of over 2 million shekel (approximately $500,000 USD) due to the software's biased VAT recalculations.

Indirectly caused: Har Shemesh farm incurred unexplained tax fines and legal costs to challenge the Tax Authority's automated decision.

Inferred additional harm: Other taxpayers likely suffered unquantified financial losses due to the software's biased calculations and unexplained fines before the issues were legally challenged.

Differential treatment

Reported: The report explicitly describes differential treatment of taxpayers caused directly by the software's biased calculations.

Directly caused: The software recalculated delayed VAT payments only when the consumer price index went up, but not when it went down, systematically disadvantaging taxpayers.

Civil rights

Reported: The report explicitly describes concerns regarding the violation of civil rights, specifically the right to information and public oversight of government actions.

Directly caused: The Tax Authority's refusal to disclose the software's logic or source code prevented the farm from understanding and challenging the administrative decisions made against them.

Indirectly caused: The delegation of administrative discretion to opaque software without public discussion undermines the public's civil right to oversee government activities.

Inferred additional harm: Widespread deployment of such opaque systems across other government agencies could systematically erode citizens' rights to due process and transparency.

People affected

• Occurrences reported: 2
• People reportedly harmed: 2
• People reportedly exposed: 2

Potential causes

Management

• Delegation of Discretion: Management allowed software to replace human judgment without oversight.
• IP Protection Prioritization: Management prioritized proprietary rights over public accountability.
• Inadequate Risk Planning: Failed to assess transparency and security risks before deployment.

Technology

• Opaque Algorithmic Logic: The software calculated fines automatically without explaining the math.
• Proprietary Executable Delivery: The software was provided as an .exe file, preventing code inspection.
• Asymmetric Calculation Logic: Fines only adjusted when the consumer price index went up, not down.

Human Factors

• Automation Bias: Tax officials relied entirely on the software output without verification.
• Lack of Technical Literacy: Employees and the public could not interpret the underlying software code.

Process and Methods

• Lack of System Documentation: No written guidelines existed for the automated calculation logic.
• No Routine Code Auditing: No process was established to review and verify software decisions.
• Inefficient Extraction Methods: Extracting guidelines required costly and complex reverse engineering.

Regulatory Environment

• Ambiguous Information Definition: Unclear legal framework on whether source code counts as public information.
• Lack of Algorithmic Mandates: No laws requiring explainability for automated government decisions.

Information quality

• Classification confidence: High
• Reason for confidence: The report provides clear, detailed accounts of the legal proceedings, the arguments of both the Tax Authority and the plaintiffs, and the specific algorithmic bias identified in the VAT calculations. There is minimal ambiguity regarding the core issues of transparency and automated decision-making.
• Ambiguities identified: The exact technical specifications and algorithms of the Tax Authority's software are not detailed, which is the central point of the legal dispute itself.
• Alternative interpretations: None, as the facts of the legal challenge and the software's lack of transparency are undisputed.

An opaque, automated tax system in Israel issued biased fines and VAT calculations, leading to legal challenges over algorithmic transparency. While the incident highlights significant issues in public sector AI governance, due process, and civil rights, its direct national security impact remains minor.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: Israel
• Alleged perpetrator: Israeli Tax Authority

Threat characteristics

• Imminence: Long-term. This represents a systemic, ongoing regulatory and legal challenge regarding public sector AI governance rather than an active, time-sensitive crisis.
• Autonomy: Full autonomy. The software independently calculated and issued tax fines and VAT adjustments without human intervention or oversight.
• Novelty: Evolved capability. While automated administrative errors are known, this case represents an evolved challenge regarding the legal precedent of disclosing proprietary government software.

Impact by dimension

• Physical security: Negligible. The incident involves administrative tax software and has no impact on physical systems, critical infrastructure, or kinetic safety.
• Information security: Negligible. No information warfare, deepfakes, or intelligence compromise occurred, though the government cited state security to avoid disclosing source code.
• Sovereignty: Minor. Automated decision-making without human oversight compromised administrative transparency and prompted supreme court-level reviews of government software disclosure policies.
• Economic security: Negligible. While taxpayers suffered financial discrepancies and the state paid a 2 million shekel settlement, there was no threat to national economic stability or strategic industries.
• Societal stability: Minor. The opaque and biased algorithmic calculations systematically disadvantaged taxpayers, raising concerns about due process, civil rights, and public oversight of government actions.