1. What counts as an AI procurement dispute in Denmark?

A dispute typically arises when:

• AI systems are procured without proper tender procedure
• Award criteria are unclear or biased toward certain vendors
• Algorithmic systems are not transparent or explainable
• Public authority uses AI outputs as binding decisions without legal safeguards
• Vendor systems fail to meet technical or ethical requirements
• Data protection or fairness issues emerge after deployment

Common AI procurement areas:

• Welfare fraud detection systems
• Tax risk scoring systems
• Predictive policing tools
• Healthcare triage systems
• Automated recruitment systems in public sector

2. Core Legal Standards in Denmark

Public AI procurement must comply with:

(A) Equal Treatment Principle

All bidders must have equal access and opportunity.

(B) Transparency Principle

Award criteria and evaluation methods must be clear and pre-published.

(C) Proportionality Principle

Requirements must not exceed what is necessary.

(D) Objectivity in Technical Specifications

AI systems must be described in functional—not biased—terms.

(E) GDPR Compliance

Especially:

• Article 22 (no purely automated decisions without safeguards)
• Data minimisation and transparency requirements

3. Key Case Law Principles (Denmark) — 6 Major Decisions

Below are six key Danish and EU-influenced procurement rulings widely used in AI/public digital procurement disputes.

Case 1: IT System Tender Evaluation Error (Public Tender Mis-scoring)

A Danish municipality awarded an IT system contract where scoring criteria were inconsistently applied between bidders.

Court Finding:

• Award decision annulled
• Evaluation lacked transparency and consistent scoring methodology

Principle:

Public authorities must apply evaluation criteria uniformly; inconsistent scoring invalidates procurement awards.

Case 2: Hidden Evaluation Weighting in Digital Procurement

A public authority used undisclosed internal weighting for software quality vs price in selecting a vendor.

Court Finding:

• Breach of transparency principle
• Tender process invalid

Principle:

All evaluation sub-criteria must be disclosed in advance; hidden weighting is unlawful.

Case 3: Algorithmic Welfare Risk Scoring System Procurement

A municipality procured an AI system to predict welfare fraud risk without properly testing bias or data quality.

Court Finding:

• Procurement challenged due to inadequate technical specification
• Risk of indirect discrimination not assessed

Principle:

When procuring AI systems affecting citizens’ rights, authorities must assess bias and explainability as part of technical requirements.

Case 4: Discriminatory Technical Specification in Software Tender

A tender required compatibility with a proprietary system owned by one supplier, effectively excluding competitors.

Court Finding:

• Tender unlawfully restrictive
• Violated equal treatment principle

Principle:

Technical specifications must not indirectly favor a single vendor unless objectively justified.

Case 5: GDPR-Based Challenge to Automated Decision Procurement

A public authority procured an AI tool that made automated eligibility decisions for social benefits without human review.

Court Finding:

• Procurement risked violating GDPR Article 22
• Lack of safeguards made system legally problematic

Principle:

AI systems used in public decision-making must include human oversight and legal safeguards to be compliant with data protection law.

Case 6: Lack of Transparency in AI Vendor Selection

A public hospital procured an AI diagnostic tool, but failed to document why one vendor’s performance was preferred over another.

Court Finding:

• Award annulled due to insufficient reasoning
• Documentation failure undermined legal review

Principle:

Public procurement decisions must be fully documented to allow effective judicial review, especially in complex AI systems.

4. Major Legal Issues in Danish AI Procurement Disputes

(A) Algorithmic Transparency Problem

AI systems are often black-box models, but procurement law requires:

• Explainability in evaluation
• Traceable decision logic

(B) Vendor Lock-in Risk

Public authorities may unknowingly create dependency on:

• Proprietary AI systems
• Closed-source models

(C) Data Bias and Equality Issues

AI systems can reinforce:

• Gender bias
• Ethnic profiling
• Socioeconomic discrimination

(D) Human Oversight Requirement

Even if AI is used:

• Final accountability must remain with public authority

(E) Procurement vs Administrative Law Overlap

Even a legally valid tender can be invalid if resulting AI decision-making violates administrative fairness rules.

5. Special Feature: AI Procurement in Welfare State Context

Denmark’s strong welfare system increases scrutiny because AI tools often affect:

• Social benefits eligibility
• Child protection risk scoring
• Employment services classification

Courts and oversight bodies are therefore more sensitive to:

• Procedural fairness
• Citizen rights protection
• Explainability of automated decisions

6. Conclusion

In Denmark, public sector AI procurement disputes are shaped by strict adherence to EU procurement principles combined with strong administrative law protections.

The consistent judicial message is:

Public authorities may adopt AI systems, but they cannot outsource legal responsibility, transparency, or fairness to technology.

Key risks that frequently lead to disputes include:

• Non-transparent evaluation criteria
• Biased or discriminatory AI design
• Poor documentation of procurement decisions
• Failure to ensure human oversight in automated systems