South Korea’s Financial Security Institute has completed what it described as the country’s first dedicated framework for testing the reliability and safety of artificial intelligence used in financial services, as banks and other firms expand AI into core operations.
According to Asia Business Daily, the institute confirmed on the 12th that the new evaluation system was built to identify risks tied to AI deployment, including hallucinations, system malfunctions, and security breaches. The move follows wider AI adoption in finance after the loosening of network separation rules.
The framework is designed around two broad pillars - reliability and safety - with 10 criteria in total. On the reliability side, the institute plans to assess model performance management, data quality, fairness and bias, and explainability. That includes checking whether performance thresholds are set appropriately, whether hallucinations and performance decay are continuously monitored, and whether the data used by models is accurate, complete, and consistent.
It also examines how bias is managed during live operations, whether customers receive clear explanations for AI-driven decisions, and whether people have channels to challenge those decisions or seek redress.
The safety side covers six areas: threats specific to AI systems, detection and response to AI-targeted attacks, protection of AI assets, vetting of external models and data, scalability of security governance, and ongoing security verification. In practice, the institute said this means testing whether adversarial attacks are detected and blocked through input screening, whether risks linked to AI models and open-source components are managed properly, and whether externally sourced models and data are checked for safety.
The framework also includes supply chain security, protections against internal data leakage, and compliance with rules for cross-border data transfers.
To build the system, the Financial Security Institute drew on South Korean and overseas standards. These include the Financial Services Commission’s AI guidelines for the financial industry, the Financial Supervisory Service’s AI risk management framework, South Korea’s AI Basic Act, ISO/IEC 42001, and Inspect, the evaluation tool developed by the UK’s AI Safety Institute.
The rollout is set to begin with an online briefing for financial firms on 14 August, followed by a demand survey in September. Pilot testing is scheduled for the second half of the year, with full evaluations due to start in 2027. The initial reviews will focus on the institute’s member companies before any broader expansion to the wider financial sector.
The institute is also considering whether the framework could later serve as a formal certification system for AI safety and reliability under the AI Basic Act.
