Do fintech companies need to explain AI credit decisions to borrowers?

Yes. Article 13 requires that AI credit scoring systems produce outputs that deployers can interpret, and CRD IV already requires explanation of adverse credit decisions. When AI drives the decision, the explanation must include the specific algorithmic factors — not just "based on our assessment criteria" but which data points and model features contributed to the outcome.

What does Article 13 require for credit scoring AI?

Instructions for use must include: the provider's identity, the system's intended purpose and creditworthiness assessment methodology, score interpretation guidance (what scores mean and how they map to credit decisions), performance metrics (accuracy, calibration, robustness) disaggregated by customer segments, known limitations, and human oversight measures for when loan officers should override the AI.

What is the compliance deadline for fintech AI transparency?

High-risk fintech AI systems (credit scoring, insurance risk assessment) must comply by August 2, 2026. The Digital Omnibus proposal (March 2026) may delay high-risk obligations to December 2027, but this has not been formally adopted. Financial institutions should plan for August 2026 and adjust if the delay is confirmed.

Does Article 13 require disclosing the credit scoring model itself?

Article 13 does not require disclosing proprietary model architecture or source code. It requires transparency about the system's characteristics, capabilities, limitations, and performance — enough for deployers to use it appropriately and for affected persons to understand the factors driving decisions about them. Trade secrets are protected under Article 78, but this does not exempt the obligation to explain outputs.

Article 13 Transparency for Fintech AI — EU AI Act

How EU AI Act Article 13 transparency applies to financial services AI — credit decision explainability, adverse action notices, borrower rights.

Published April 2, 2026

Article 13 of the EU AI Act requires high-risk AI systems to be sufficiently transparent for deployers to interpret outputs and use them appropriately. In financial services, transparency directly affects borrowers' rights, investor protection, and systemic risk management. Under Annex III, Section 5(b), AI systems used for creditworthiness assessment and credit scoring are classified as high-risk and subject to the full scope of Article 13 transparency obligations.

This page explains how Article 13 transparency requirements apply specifically to fintech — from credit decisions to algorithmic trading — including borrower explanation rights, instructions for loan officers, adverse action transparency, and what code-level violations Scanara detects in financial AI systems.

Which Fintech AI Systems Must Be Transparent?

Credit Scoring and Lending Decisions

AI systems assessing creditworthiness or making lending recommendations must explain which factors drove the decision. “Your application was declined by our automated system” without specific reasons is non-compliant under Article 13(1).

Insurance Risk Assessment

AI pricing and underwriting systems that determine premiums or coverage eligibility must be transparent about the factors influencing risk assessments. Policyholders need to understand why their premium was set at a particular level.

Anti-Money Laundering (AML) Screening

AI systems flagging suspicious transactions or customers for AML review must provide compliance officers with interpretable outputs explaining why a transaction or account was flagged, enabling appropriate human investigation.

Investment Suitability and Robo-Advisory

AI systems recommending investment products must explain the rationale behind recommendations. Under MiFID II, suitability assessments must be transparent — Article 13 adds system-level transparency about how the AI reaches its recommendations.

Fraud Detection Systems

AI flagging potentially fraudulent transactions must provide fraud analysts with interpretable alert context. A fraud score without explanation of contributing signals prevents appropriate investigation as required by Article 13(1).

Key Transparency Obligations for Fintech

Article 13 requires instructions for use that enable deployers to interpret AI outputs appropriately. For fintech, this means loan officers and compliance staff must understand how AI scores map to creditworthiness, what the system's known limitations are, and when human judgment should override algorithmic recommendations.

Borrowers Must Receive Explanations

When a loan is declined based on AI, the borrower must receive an explanation of which factors drove the decision. Article 13 combined with CRD IV Article 147 requires that adverse credit decisions include specific reasons, not just a generic rejection notice. The explanation must reference the algorithmic factors that contributed to the outcome.

AI Score-to-Decision Mapping Must Be Documented

Instructions for use must explain to loan officers how the AI score maps to creditworthiness and its known limitations. A raw score (“Credit risk: 0.73”) without context on what that means for lending decisions, how it was derived, and when it may be unreliable prevents appropriate use under Article 13(1).

Adverse Action Notices Must Include AI Factors

Adverse action notices (already required under CRD) must now include AI-specific transparency about algorithmic factors. Generic statements like “based on our assessment criteria” are insufficient when the criteria are algorithmically determined — the specific AI-driven factors must be disclosed.

Performance Metrics Must Be Disclosed

Article 13(3)(b)(ii) requires disclosure of accuracy, robustness, and cybersecurity performance. For credit scoring, this means reporting default prediction accuracy, calibration metrics, and known performance variations across customer segments — disaggregated to reveal potential disparate impact.

Regulatory Overlaps

Fintech AI transparency sits at the intersection of financial regulation, consumer protection, and data protection law. Article 13 must be coordinated with sector-specific transparency requirements already in place.

EBA Guidelines on AI in Credit Assessment

The European Banking Authority has issued guidance on AI use in creditworthiness assessment, requiring explainability and transparency for AI-driven credit decisions. These guidelines align with Article 13 but add sector-specific requirements around model validation, documentation, and supervisory reporting that financial institutions must meet alongside AI Act obligations.

CRD IV Article 147 — Credit Decision Explanation

The Capital Requirements Directive requires institutions to provide reasons for adverse credit decisions. Article 13 extends this by requiring transparency about the AI system itself — not just the decision outcome, but the system's characteristics, capabilities, limitations, and performance metrics that underpin the decision process.

MiFID II — Suitability and Appropriateness

MiFID II requires investment firms to assess client suitability and provide suitable recommendations. When AI drives suitability assessments, both MiFID II transparency (to clients) and Article 13 transparency (to deployers) must be satisfied. Advisors must understand the AI logic to fulfil their MiFID II obligations.

PSD2 — Payment Services Transparency

The Payment Services Directive requires transparency in payment processing, including when transactions are declined. AI-driven fraud detection or transaction screening must provide meaningful explanations when payments are blocked — Article 13 adds requirements for deployer-facing documentation of the screening system's logic and limitations.

Common Violations Scanara Detects

Credit Rejection Without Algorithmic Factor Disclosure

Loan decision endpoints returning rejection responses without including the specific algorithmic factors that drove the adverse decision. Article 13(1) requires interpretable outputs — Scanara flags credit decision APIs that return only a status code without factor-level explanation.

Risk Score Without Score-to-Decision Mapping

API responses returning credit risk scores without documentation mapping scores to creditworthiness categories. A raw probability or score without context on thresholds, interpretation, and confidence intervals prevents loan officers from using the output appropriately.

Missing Disclosure Header in Customer-Facing API Responses

Customer-facing endpoints (account dashboards, loan portals) that display AI-driven assessments without disclosing that the assessment was generated by an AI system. Article 50(1) requires notification when natural persons interact with AI systems.

Article 13 Compliance Checklist for Fintech

Art. 13(1)

Make Credit Decisions Interpretable

Ensure loan officers and compliance staff can understand which factors drove an AI credit recommendation. Provide factor-level explanations, score interpretation guidance, and confidence indicators alongside every AI output.

Art. 13(3)(b)

Document Score-to-Decision Mapping

Provide instructions explaining how AI scores map to creditworthiness, what score ranges mean for lending decisions, and when loan officers should exercise independent judgment. Include thresholds, confidence intervals, and edge case guidance.

CRD IV

Include AI Factors in Adverse Action Notices

Ensure adverse credit decision notices include the specific algorithmic factors that contributed to the decline. Generic rejection language must be augmented with AI-specific reasons that satisfy both CRD IV and Article 13 requirements.

Art. 13(3)(b)(ii)

Disclose Performance and Accuracy Metrics

Report model accuracy, calibration, and robustness metrics disaggregated by relevant customer segments. Include known performance limitations, validation methodology, and conditions under which the model's predictions may be less reliable.

Art. 50(1)

Notify Customers of AI Use

Inform borrowers and customers when AI is involved in decisions affecting them. Disclosure should occur before or at the point of the AI-driven decision in customer-facing portals, loan applications, and account dashboards.

Frequently Asked Questions

How Scanara Helps

Scanara automates EU AI Act compliance from code to dossier. Connect your GitHub repos and get compliance reports in minutes.