TLDR: Commercial lenders are facing significant challenges in fully adopting generative AI technologies due to existing regulations that prohibit the use of unencrypted personal credit information and the persistent risk of AI “hallucinations.” This is delaying the full potential of AI in enhancing customer service and operational efficiency within the financial sector.
Commercial lenders are encountering substantial difficulties in the complete integration of generative artificial intelligence (AI) technologies into their financial services. A primary impediment is the current regulatory framework, which prohibits the use of full, unencrypted personal credit information. This regulatory constraint, coupled with the lingering risk of “hallucinations” – instances where AI generates inaccurate or nonsensical outputs – is rendering vast amounts of rich datasets unusable, despite their potential value.
Local banks have begun to embrace generative AI for applications such as credit risk assessment and assisting employees in customer consultations. However, the full realization of these technologies is being hampered by the aforementioned challenges. An official from Woori Financial Group highlighted that “Generative AI adoption is a paradigm change,” emphasizing that comprehensive deregulation would significantly enhance customer service and workflow efficiency, even while acknowledging the unresolved cybersecurity and privacy risks. The official also noted that less experienced employees could see a reduction in the time required for customer assistance, thereby boosting productivity, and anticipated that “more AI-mediated services to boost our overall support capabilities.”
This situation unfolds amidst delays in regulatory changes concerning network separation within the financial sector. The Financial Services Commission (FSC) is reportedly postponing its initial plans to permit the use of unencrypted data in generative AI-mediated financial services until next year, pushing back from an earlier target of the latter half of this year. This cautious approach by regulators is driven by the need to strike a balance between fostering innovation and ensuring security in network separation. Under current regulations, financial service providers must secure regulatory sandbox approval from the FSC to deploy generative AI services within their internal networks, and even with such approval, only encrypted data is permitted for use.
Also Read:
- Cybersecurity Hurdles Slow Generative AI Adoption in South Korean Banking Sector
- Financial Giants JPMorgan and Robeco Quietly Integrate AI into Core Operations
An industry official, speaking anonymously, acknowledged the financial authorities’ cautious stance, citing concerns about potential fallout from a lack of regulatory fallback. However, the official also warned that “delays in technology implementation will inevitably lead to delays in service quality improvements overall.” As of last December, the FSC had approved a pilot program allowing nine financial service providers to use encrypted personal data for generative AI services within their internal networks. Current AI-powered financial services are designed to comply with privacy and data protection regulations, primarily assisting bank employees with extensive data analysis. For instance, Woori Bank introduced an AI knowledge consultation system in April, utilizing agentic retrieval-augmented generation. This system analyzes user inquiries, sets goals, and gathers and evaluates information from multiple sources, enabling sales employees to provide accurate, real-time information on financial regulations, procedural protocols, and product details.
