Core banking platforms operate under stringent availability, integrity, and regulatory expectations while evolving rapidly through continuous delivery, vendor updates, and complex integration landscapes. Traditional quality assurance approaches struggle to maintain defect containment, security assurance, and audit readiness as test suites scale and system dependencies grow. This paper proposes a risk-aware AI framework that continuously converts operational, cybersecurity, compliance, and model risks into test strategy decisions, enabling automated test selection, prioritization, and governance aligned with core banking constraints. The framework integrates (i) risk-driven change impact analysis, (ii) AI-based defect propensity and failure propagation signals, (iii) policy-aware test orchestration with verifiable control mapping, and (iv) model governance primitives for explainability, robustness, and bias monitoring. A simulation-based evaluation illustrates how risk-weighted prioritization improves time-to-detection and risk-weighted recall under fixed execution budgets. The approach is designed for hybrid stacks typical of banks (batch and online, legacy and microservices) and emphasizes audit-friendly evidence generation, reducing unknown-risk exposure while improving release velocity.