South Korea’s Financial Services Commission has released revised self-regulatory AI guidelines effective June 22, 2026, applicable to all financial companies and fintech businesses — a move that reshapes how Korean insurers may deploy AI in underwriting, claims, fraud detection, and distribution. The update arrives as global standard-setters converge on similar principles, placing Seoul in alignment with an emerging international consensus on responsible AI adoption in finance.
Seven Principles and a Hard Line on Human Oversight
The FSC AI guidelines establish seven core principles: Governance, Legitimacy, Means of Assistance (human supervision), Credibility, Financial Stability, Good Faith, and Security. The architecture is deliberate: rather than a prescriptive technical rulebook, the FSC has opted for a principles-based framework that places responsibility squarely at board level. CEOs and senior executives of financial companies are expected to lead AI oversight, not delegate it downward.
The most operationally consequential principle for insurers is the third: AI must be used as a means of assistance, with humans retaining ultimate decision-making authority. In practice, this means automated underwriting engines and AI-driven claims triage systems cannot operate as final arbiters — human review must remain in the loop for decisions that materially affect policyholders. This constraint will force a re-examination of straight-through processing workflows at carriers that have moved furthest toward automation, echoing the human-in-the-loop requirements that APRA called for in its own step-change in AI risk governance across the insurance sector.
The framework draws a material distinction between standard and high-impact AI. High-impact AI subject to South Korea’s Framework Act on AI Development must follow mandatory regulatory requirements, while other AI use may apply the guidelines on a discretionary basis. South Korea’s AI Basic Act took effect on January 22, 2026, establishing baseline obligations for high-impact AI including loan screening — with underwriting and credit-scoring for insurance likely to fall under similar scrutiny. The maximum administrative fine under the AI Basic Act is KRW 30 million, deliberately lower than EU AI Act penalties to reflect a pro-innovation posture. For large Korean insurers, the fine itself is immaterial; the reputational and supervisory-relationship risk of non-compliance is the real deterrent.
Network Separation Reform and the Phased Cybersecurity Sandbox
Alongside the governance principles, the FSC is unwinding one of Korea’s most restrictive IT infrastructure rules to enable AI deployment. The network separation rule easing applies to 49 financial companies with at least KRW 10 trillion in assets and a regular staff of 1,000 or more — a threshold that captures all of Korea’s major insurance groups alongside the large banks and securities firms. These institutions will be required to appoint a Chief Information Security Officer under the Electronic Financial Transactions Act.
The implementation proceeds in three phases: Phase 1 (June–July 2026), Phase 2 (August–September 2026), and Phase 3 (Q4 2026). An FSC AI cybersecurity sandbox will begin testing operations in the second half of 2026 for qualified financial entities with advanced AI capabilities. The phased rollout gives large insurers a structured on-ramp but also compresses timelines: compliance teams must map AI systems against the new framework while simultaneously preparing for the sandbox entry criteria. For insurtech players and smaller carriers below the asset threshold, the discretionary application of guidelines provides breathing room — though the direction of regulatory travel is unambiguous.
The Fraud Mandate: 9 Trillion Won of Exposure Driving the AI Imperative
The governance update cannot be read in isolation from a parallel initiative that makes the AI case more urgent. Korean insurers paid out 1.16 trillion won in fraudulent claims in 2025 — a record high, up 6.9 billion won from 2024. That confirmed figure substantially understates the problem: the FSC estimates total insurance fraud exposure in Korea at approximately 9 trillion won, nearly eight times the confirmed payout figure. The gap between detected and estimated fraud is a standing indictment of legacy detection systems.
The breakdown by line reveals where AI-powered fraud detection investment is most urgent. Long-term non-life insurance accounts for the largest share of fraudulent payouts at 44.7%, followed by auto insurance at 22.4%, life insurance at 21.8%, and general non-life at 11.2%. Long-term non-life — health-linked products with complex benefit structures and high claim frequency — is precisely where machine-learning pattern recognition has the highest potential return.
The FSC launched an AI-Based Insurance Fraud Prevention System Task Force on June 4, 2026, with a platform development plan deadline of September 2026 and legislative follow-through beginning October 2026. The dual mandate embedded in the revised AI guidelines — govern AI responsibly while deploying it aggressively against fraud — creates a structural tension that Korean insurers must manage: the same human-oversight principle that constrains automated underwriting must be calibrated carefully so it does not blunt the speed advantage of real-time fraud scoring. This dynamic mirrors the challenge facing Sixfold’s AI underwriter as it tests the limits of point-solution automation in a human-in-the-loop world.
Global Convergence: FSB Sound Practices and the IAIS Application Paper
Korea’s updated framework does not emerge from a regulatory vacuum. Two multilateral processes are shaping the international backdrop against which FSC’s move should be assessed.
The Financial Stability Board published a consultation report on June 10, 2026, proposing 12 sound practices for responsible AI adoption, covering organisation-wide governance (practices 1–4), AI lifecycle risk management (practices 5–10), and cyber, IT, and third-party risk (practices 11–12). The FSB consultation closes on July 22, 2026, with a final report expected in October 2026. The FSB explicitly notes that recent developments in frontier AI models highlight the dynamic nature of this technology and the rapid pace at which its capability evolves — a recognition that governance frameworks must be adaptive rather than static. The FSB consultation is open for comment at fsb.org, where the full 12-practice consultation document is available.
At the insurance-specific level, the IAIS Application Paper on AI supervision (July 2025) covers five areas: risk-based supervision and proportionality, governance and accountability, robustness, safety and security, transparency and explainability, and fairness, ethics, and redress. The IAIS application paper reinforces how existing Insurance Core Principles remain essential considerations for supervisors and insurers navigating AI adoption — a continuity argument that gives Korean supervisors a ready-made international reference for their own principles-based approach. Korea’s seven principles map closely onto the IAIS five-area taxonomy: Governance aligns with governance and accountability; Credibility with transparency and explainability; Good Faith with fairness and ethics. The structural similarity is not coincidental — it reflects deliberate design at the FSC to ensure compatibility with international supervisory expectations, a positioning advantage for Korean insurers operating cross-border, as illustrated by DB Insurance’s acquisition of Fortegra as Korean insurers target US specialty scale.
The convergence of FSC, FSB, and IAIS frameworks also affects capital and investment flows. AI-labeled insurtechs’ record share of global insurtech funding — investors pricing in governance-compliant AI platforms as a premium asset class. Korean carriers that demonstrate early, credible AI governance alignment will find it easier to partner with, or acquire, such platforms under the FSC’s evolving regulatory architecture. The full FSC revised AI guidelines are published on fsc.go.kr with the effective date and seven principles detailed.
What Korean Insurers Must Act on Now
For compliance and technology officers at Korean insurance groups, the June 22 effective date compresses the action window. Three priorities stand out. First, AI inventory: carriers need a clear map of which systems constitute high-impact AI under the Framework Act versus discretionary-guidelines AI, since the mandatory versus voluntary distinction determines compliance exposure. Second, governance structure: the board-level accountability requirement means governance frameworks cannot remain in IT or actuarial silos — risk and compliance functions must formalize escalation paths for AI-related decisions affecting policyholders. Third, the fraud detection build: with the platform development plan due in September 2026 and legislation following in October, the window for insurers to shape technical standards through the Task Force is narrow. Early engagement with the FSC on fraud-detection architecture — including how human-oversight requirements apply to automated fraud flags — will determine whether the new rules enable or constrain AI-driven loss ratio improvement.
