Artificial Intelligence in Fintech and the Black Box Dilemma: Regulatory Perspectives in Azerbaijani Law (OpenAlex (recherche mondiale))

Summary

• Artificial intelligence adoption in fintech enhances operational speed in credit scoring, insurance risk assessment, and fraud detection, but creates algorithmic opacity known as the black box dilemma.p.1
• Modern AI applications in financial services span four primary areas: fraud detection in digital payments, virtual assistant services in banking, algorithmic trading, and regulatory compliance.p.6
• The European Union's 2024 Artificial Intelligence Act establishes a risk-based regulatory framework classifying AI systems into four risk tiers: minimal, limited, high, and unacceptable.p.12
• Under GDPR Article 22.1, which entered into force in 2018, data subjects have the right not to be subject to decisions based solely on automated processing.p.19
• The Court of Justice of the European Union ruled on 7 dekabr 2023 in the SCHUFA case that automated credit scoring is subject to GDPR restrictions when it plays a decisive role in credit granting.p.21
• The Central Bank of Azerbaijan introduced a SupTech Roadmap covering 2024-2029 to enhance supervisory technology and facilitate human control over AI systems in finance.p.22
• Azerbaijan's AI Strategy for 2025–2028 aims to establish legal, ethical, and normative frameworks to mitigate algorithmic bias and data privacy risks.p.26

Key insights

1. Deep learning algorithms continuously adapt through trial and error, making their internal decision logic opaque to both users and financial regulators.p.3
2. Algorithmic bias in AI credit scoring can exacerbate financial exclusion by assigning lower scores to specific demographic groups based on historical data.p.10
3. Under the EU AI Act, AI systems used for assessing individual creditworthiness or setting premiums for life and health insurance are explicitly classified as high-risk.p.13
4. Explainable AI methods provide crucial insight into algorithmic outcomes but are insufficient on their own without meaningful human oversight.p.18
5. Meaningful human control requires that human supervisors possess the requisite competence, training, authority, and power to override automated AI decisions.p.19
6. In the landmark SCHUFA case judgment on 7 dekabr 2023, the CJEU held that automated third-party credit scoring constitutes an automated decision under GDPR Article 22.1.p.21
7. Azerbaijan's Central Bank utilizes a special regulatory sandbox regime under Article 50.1 of its organic law to test fintech AI products before public deployment.p.15
8. Implementing a human-in-the-loop model combines machine computational power with human intuition and ethical judgment to minimize operational and legal risks.p.23