Financial Regulation and AI: Compliance and Ethics

Module: Fintech AI | Difficulty: Advanced

Regulatory Framework

• GDPR: Data protection
• ECOA: Fair lending
• SR 11-7: Model risk
• EU AI Act: AI regulation

Fair Lending

Algorithmic Accountability

Bias Detection

import numpy as np

class FairnessChecker:
    def __init__(self, predictions, sensitive_attributes):
        self.preds = predictions; self.attrs = sensitive_attributes
    def demographic_parity(self):
        groups = np.unique(self.attrs)
        rates = [self.preds[self.attrs == g].mean() for g in groups]
        return max(rates) - min(rates)
    def equalized_odds(self, true_labels):
        groups = np.unique(self.attrs)
        tpr_diffs = []; fpr_diffs = []
        for g in groups:
            mask = self.attrs == g
            tpr = self.preds[mask & (true_labels == 1)].mean()
            fpr = self.preds[mask & (true_labels == 0)].mean()
            tpr_diffs.append(tpr); fpr_diffs.append(fpr)
        return max(tpr_diffs) - min(tpr_diffs), max(fpr_diffs) - min(fpr_diffs)

Research Insight: AI in finance faces increasing regulatory scrutiny. The key challenge is balancing model performance with fairness and explainability. Simple models (logistic regression) are often preferred over complex ones (neural networks) because they are easier to explain and audit.