黑箱決策阻礙問責追溯

Risk Description

Deep learning model opacity makes it difficult to trace decision causes and accountability when AI systems cause harm. Affected individuals cannot obtain meaningful explanations to understand and challenge adverse AI decisions. This directly violates many regulations (e.g., GDPR Article 22) requiring explanation rights for automated decisions, and prevents organizations from identifying and correcting root causes of adverse outcomes.