AI in Banking and Insurance Operations — 2026-09-02

Top developments

US Agencies Clarify Model Risk Management for Generative AI

The Office of the Comptroller of the Currency (OCC), in coordination with the Federal Reserve and FDIC, has issued revised guidance on Model Risk Management. The updated guidance explicitly clarifies that traditional model risk management principles do not apply to generative or agentic AI systems, reflecting a shift toward risk-based approaches tailored to specific operational complexities. This distinction is critical for banks deploying Large Language Models (LLMs) for customer service or internal coding, separating them from traditional credit scoring models.

EU Supervisory Authorities Demand Consistent Frontier AI Oversight

The European Banking Authority (EBA), EIOPA, and ESMA published a joint statement calling for a cross-sectoral, risk-based supervisory approach to mitigate ICT risks stemming from frontier AI models. This move aims to prevent regulatory arbitrage across banking, insurance, and securities sectors within the EU, ensuring that the rapid integration of advanced AI does not outpace governance frameworks.

South Korea Launches Public Campaign Against AI Investment Fraud

The Financial Services Commission and other Korean financial authorities have initiated a nationwide public relations campaign to combat illegal financial crimes utilizing AI. These schemes involve impersonating experts via deepfakes and spreading fake news promising high returns with principal guarantees. This proactive stance highlights the growing threat of AI-generated social engineering attacks targeting retail investors.

KB Financial Group Breaks "Network Separation" Barriers with GenAI

KB Securities and KB Kookmin Bank showcased a practical generative AI architecture that overcomes South Korea’s strict network separation regulations and cloud operation complexities. Announced at the AWS Financial Services Forum on August 27, this implementation demonstrates how major Korean banks can deploy secure, compliant GenAI solutions despite rigorous local data security laws.

Insurers Accelerate GenAI for Claims and Underwriting

The insurance sector is seeing a rapid shift from pilot programs to production-scale AI deployments. Industry reports indicate that AI is now a core driver of productivity and customer experience expansion, though firms are urged to address emerging risk management gaps. Simultaneously, security vendors like Yatab are deploying AI-DLP (Data Loss Prevention) systems for top-tier Korean financial institutions to monitor and block sensitive data leakage during generative AI interactions.

Local view

South Korean media outlets are closely monitoring the dual narrative of innovation and security. EBN reports that while AI adoption in the insurance sector is accelerating rapidly to improve productivity, there are growing concerns about whether risk management strategies are keeping pace with this technological speed.

Asia Economy highlights the government's urgent response to AI-fueled scams, emphasizing the need for public education to counter sophisticated deepfake and fake news tactics used by illegal investment brokers.

Venture Square details the rise of specialized security solutions, noting that Yatab has secured contracts with first-tier financial institutions to implement AI-DLP systems that detect personal and credit information leakage in real-time during generative AI usage.

Context & numbers

• Fraud Prevention ROI: JPMorgan’s AI fraud system is estimated to prevent $1.5 billion in losses annually with a 98% accuracy rate, illustrating the high stakes for model accuracy.
• Adoption Rates: Approximately 90% of banks now use AI in some capacity, driving a $36.6 billion revolution in finance operations.
• Savings Disclosed: 42% of card issuers reported saving more than $5 million in fraud losses through AI-enabled Decision Intelligence systems.

On the radar

• QA Financial September Roundup: Regulators in the UK and Europe are expected to release further compliance news in early September, with a consistent message focusing on operational resilience and AI governance.
• Goldman Sachs AI Agent Integration: Engineers at Goldman Sachs are actively working to transfer "tribal knowledge" into AI agents, signaling a deeper integration of agentic workflows into core investment banking processes.