South Korea regulates AI in banking on two tracks. The Financial Services Commission's AI guidelines for the financial sector, announced on 18 June 2026 and in effect since 22 June 2026, set seven principles (governance, legitimacy, AI as a means of assistance, credibility, financial stability, good faith and security) and are self-regulatory, with the Financial Supervisory Service's AI risk management framework spelling out the governance principle. Binding duties come from the Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust (Act No. 20676), in force since 22 January 2026: it lists AI used for judgments that significantly affect individuals' rights and obligations, such as loan screening, as high-impact AI with risk management, explanation, user protection and human oversight duties, and it fines a failure to give advance notice of high-impact or generative AI up to 30 million won. Since May 2026 the FSC has also eased its network separation rule so that selected financial companies can test frontier AI models for cyber defence.
| Full name | Financial Services Commission (FSC) and Financial Supervisory Service (FSS) of the Republic of Korea, with the Ministry of Science and ICT under the Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust (AI Basic Act) |
| Role | Financial regulator (FSC) and supervisor (FSS) whose AI rulebook is self-regulatory guidance, working since January 2026 beside a national AI law that puts direct duties on firms providing high-impact or generative AI, and using its binding network separation rule to control how financial companies connect to AI |
| Force on banks | Binding law |
| Applies to | Financial companies supervised by the FSC and FSS, including banks, financial holding companies, insurers, financial investment firms, card companies, credit information companies, virtual asset service providers and electronic financial service providers; the 2026 AI guidelines also reach fintech and other non-financial firms whose AI can affect financial transactions. The FSC's AI guidelines and the FSS AI risk management framework are voluntary. The AI Basic Act is binding: a financial company that provides a product or service using generative or high-impact AI, which can include AI used in loan screening, has the Act's notice and high-impact duties, enforced by the Ministry of Science and ICT |
| Key document | AI Guidelines in the Financial Sector (FSC, announced 18 June 2026, in effect 22 June 2026), read with the Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust (Act No. 20676, in force 22 January 2026) |
| Latest move | 18 Nov 2026: the FSC proposed amendments to the supervisory regulation on electronic financial services (up-to-date IT asset registers, third-party vulnerability evaluations every six months for large firms, emergency inspections on demand), with comments to 30 November 2026 and effect planned for the first quarter of 2027, following its 5 November plan against AI-driven hacking; 3 Sep 2026: second phase of the temporary easing of network separation for frontier-AI cyber testing (75 eligible companies, 15 to be selected). |
| Documents tracked | 8 · all documents → |
The FSC's guidance runs from a 2021 best-practice standard to a 2026 set of principles. Its Guideline on the Use of Artificial Intelligence in Financial Services of 8 July 2021 asked every financial sector using AI for three internal controls (AI ethics principles, a dedicated organisation and a risk management policy), for stronger controls and a designated officer where AI decisions significantly affect whether a customer's financial contract is concluded or maintained, and for notice to customers of AI use and of their right to an explanation under the Credit Information Act. An August 2022 plan to promote AI and ensure trust followed, with detailed guidelines from the Korea Federation of Banks for five services, and in April 2023 a verification system for AI credit scoring models and an AI security guideline. On 18 June 2026 the FSC announced revised guidelines, in effect from 22 June 2026, that integrate the earlier texts under seven principles; the FSS's AI risk management framework details the governance principle and the Financial Security Institute's AI security guidebook the security principle. An FSS survey of 118 banks, securities firms, insurers and card companies in April 2025 had found that only 5 banks, 4 insurers and 1 securities firm had an AI decision-making body and that about 85% had no AI ethics principles or risk management standards.
The national layer is binding. The Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust (Act No. 20676), promulgated on 21 January 2025, entered into force on 22 January 2026 with its Enforcement Decree and is administered by the Ministry of Science and ICT, not by the financial regulators. It applies to AI business operators, which include financial companies that provide AI products or services to customers, and to conduct abroad that affects the Korean market or users. A company that provides a product or service using generative or high-impact AI must notify users in advance (Article 31), and high-impact AI, which includes AI used for judgments with a significant impact on individuals' rights and obligations such as loan screening, brings duties on risk management, explanation, user protection, human oversight and documentation (Article 34). Article 34(3) treats equivalent measures taken under other laws as compliance; the FSC's guidelines say this covers the explanation duty for firms meeting Articles 35-2 and 36-2 of the Credit Information Act and the user-protection duty for sellers meeting Article 10 of the Financial Consumer Protection Act. Where the Act and the FSC's guidelines both apply, the guidelines give the Act precedence.
The FSC's sharpest tool is its network separation rule under the supervisory regulation on electronic financial services, which restricts financial companies' internet access from their internal systems. Its 13 August 2024 roadmap let them use cloud-based generative AI through regulatory sandbox exemptions with security conditions, and in December 2024 it added support for open-source models installed on internal networks. After reviewing threats from frontier models such as Mythos, it announced on 22 May 2026 one-year no-action letters letting qualified large financial companies use advanced AI models and SaaS tools for cyber defence, followed on 2 July 2026 by sanction relief for minor outages caused by AI security testing and patching and a six-area response guide, and on 3 September 2026 by a second phase open to 75 companies. On 5 November 2026, after 40 hacking incidents at 37 financial companies between 30 September and 3 November, it set out work toward a complete lifting of the rule, permanent use of AI-SaaS for cyber defence and a digital finance safety bill planned for the first quarter of 2027.
What has the South Korea (FSC / FSS) actually published on AI?
| Date | Document | Status |
|---|---|---|
| Jul 2, 2026 | Korea FSC Frontier AI Security Response Guide — Financial Sector Response Guide to Frontier AI Security Threats (프런티어 AI 보안위협 금융분야 대응요령) | Final |
| Jun 18, 2026 | Korea FSC AI Guidelines for the Financial Sector — AI Guidelines in the Financial Sector (금융분야 인공지능 가이드라인) | In force |
| May 22, 2026 | Korea FSC Frontier AI Network Separation Easing — Network Separation Rules to be Eased in Financial Sector to Boost Innovation and Cybersecurity in AI Transformation | Final |
| Jan 15, 2026 | Korea FSS AI Risk Management Framework — AI Risk Management Framework in the Financial Sector (금융분야 AI 위험관리 프레임워크, AI RMF) | In force |
| Jan 21, 2025 | Korea AI Basic Act — Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust (인공지능 발전과 신뢰 기반 조성 등에 관한 기본법) | In force |
| Aug 13, 2024 | Korea FSC Network Separation Roadmap — Roadmap to Improve Network Separation in the Financial Industry (금융분야 망분리 개선 로드맵) | Final |
| Apr 17, 2023 | Korea FSC AI Credit Scoring and AI Security Guideline — Verification System for AI-driven Credit Scoring Models and Security Guideline for the Use of AI in the Financial Sector | Final |
| Jul 8, 2021 | Korea FSC AI Guideline 2021 — Guideline on the Use of Artificial Intelligence in Financial Services (금융분야 인공지능(AI) 가이드라인) | Superseded |
| Date | Type | Document / event |
|---|---|---|
| Nov 18, 2026 | Milestone | Proposed amendments to the supervisory regulation on electronic financial services. Follow-up to the 5 November plan: all firms under the Electronic Financial Transactions Act would keep an up-to-date register of IT assets, large firms (KRW2 trillion in assets and 300 staff) would have third-party vulnerability evaluations every six months instead of yearly, and the FSC could demand emergency vulnerability inspections. Comments ran from 18 to 30 November 2026; the rules are planned to take effect in the first quarter of 2027. |
| Nov 5, 2026 | Milestone | FSC plan to strengthen the cyber defence framework against AI-driven hacking. After 40 hacking incidents at 37 financial companies between 30 September and 3 November 2026, the FSC announced inspections of the external contact points and internal controls of about 580 entities, and set out a taskforce toward a complete lifting of the network separation rule, permanent use of AI-SaaS for cyber defence, cybersecurity internal-control guidelines and a digital finance safety bill, both planned for the first quarter of 2027. |
| Sep 3, 2026 | Milestone | Second phase of the network separation easing for frontier-AI cyber testing. The FSC widened eligibility from 49 to 75 companies by lowering the threshold to KRW2 trillion in assets and 300 regular staff, which brings in nonbank financial companies, and by adding electronic financial service providers with at least KRW2 trillion in annual electronic financial transactions; the number to be selected rose from 10 to 15. Selected firms receive a one-year no-action letter and must report test findings, which the FSC says will feed updates to its AI guidelines. |
| Jul 2, 2026 | Guidance | Korea FSC Frontier AI Security Response Guide — Financial Sector Response Guide to Frontier AI Security Threats (프런티어 AI 보안위협 금융분야 대응요령). On 30 June 2026 South Korea's Financial Services Commission approved sanction relief for minor IT failures caused by AI security testing and urgent security patches, and after consulting the FSS and the Financial Security Institute it issued a response guide for financial companies facing security threats from frontier AI such as Mythos, published on 2 July 2026 in Korean. source ↗ |
| Jun 18, 2026 | Guidance | Korea FSC AI Guidelines for the Financial Sector — AI Guidelines in the Financial Sector (금융분야 인공지능 가이드라인). South Korea's Financial Services Commission announced revised AI guidelines for the financial sector on 18 June 2026, in effect from 22 June 2026, integrating its 2021 AI guideline, 2022 development and use guidebook and 2023 AI security guideline. source ↗ |
| May 22, 2026 | Guidance | Korea FSC Frontier AI Network Separation Easing — Network Separation Rules to be Eased in Financial Sector to Boost Innovation and Cybersecurity in AI Transformation. On 22 May 2026, after reviewing cyber threats from advanced AI models such as Mythos, South Korea's Financial Services Commission announced that it would immediately ease its network separation rule so that qualified financial companies can use advanced AI models to assess vulnerabilities and use SaaS security tools. source ↗ |
| Jan 15, 2026 | Framework | Korea FSS AI Risk Management Framework — AI Risk Management Framework in the Financial Sector (금융분야 AI 위험관리 프레임워크, AI RMF). South Korea's Financial Supervisory Service published its AI Risk Management Framework in the Financial Sector (AI RMF) in draft on 15 January 2026, setting out three core processes for financial companies: AI governance, AI risk assessment and AI risk control. source ↗ |
| Dec 22, 2025 | Milestone | FSC presents the direction for revising its financial AI guidelines. At a financial-sector AI council the Korea Institute of Finance presented a draft that integrates the 2021 operational guideline, the 2022 development and use guidebook and the 2023 AI security guideline under seven principles, to remain a best-practice, self-regulatory standard. The financial AI platform run by the Korea Credit Information Services opened the same day. Published in Korean. |
| Jan 21, 2025 | Statute | Korea AI Basic Act — Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust (인공지능 발전과 신뢰 기반 조성 등에 관한 기본법). South Korea's Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust (Act No. source ↗ |
| Dec 12, 2024 | Milestone | FSC announces support measures for generative AI in the financial sector. At a financial-sector AI council the FSC added support for open-source AI installed on internal networks to the internet-based commercial AI allowed under the August 2024 network separation roadmap, planned a financial AI platform and finance-specific Korean-language data, and said it would revise its AI guidelines. Published in Korean. |
| Aug 13, 2024 | Framework | Korea FSC Network Separation Roadmap — Roadmap to Improve Network Separation in the Financial Industry (금융분야 망분리 개선 로드맵). On 13 August 2024 South Korea's Financial Services Commission announced a roadmap to improve the network separation rule, which restricts financial companies' internet access, saying that cloud-based software and generative AI had made the ten-year-old rule an obstacle. source ↗ |
| Apr 17, 2023 | Guidance | Korea FSC AI Credit Scoring and AI Security Guideline — Verification System for AI-driven Credit Scoring Models and Security Guideline for the Use of AI in the Financial Sector. On 17 April 2023 South Korea's Financial Services Commission introduced a verification system for AI-driven credit scoring models and a security guideline for the use of AI in the financial sector, follow-up measures to its August 2022 plan to promote AI use and ensure trust in finance. source ↗ |
| Jul 8, 2021 | Guidance | Korea FSC AI Guideline 2021 — Guideline on the Use of Artificial Intelligence in Financial Services (금융분야 인공지능(AI) 가이드라인). South Korea's Financial Services Commission announced its Guideline on the Use of Artificial Intelligence in Financial Services on 8 July 2021, a best-practice standard for every financial sector that uses AI in financial transactions or customer services. source ↗ |
- The FSC's work from the second half of 2026 on regulatory changes for AI transformation, AI risk management and pilot tests of AI agents in the regulatory sandbox, and any rules on the responsibility and authority of AI agents that recommend, sell or pay for financial products.
- The complete lifting of the network separation rule that the FSC's November 2026 taskforce is to pursue, permanent use of AI-SaaS for cyber defence, and the schedule for a third phase of the temporary easing.
- The digital finance safety bill the FSC plans to introduce in the first quarter of 2027, the cybersecurity internal-control guidelines due in the same quarter, and the final amendments to the supervisory regulation on electronic financial services, planned to take effect in the first quarter of 2027.
- Ministry of Science and ICT notices and guidance on high-impact AI under the AI Basic Act, and any confirmation under Article 33 of whether a financial use such as loan screening is high-impact.
- Updates to the frontier-AI response guide, which the FSC says will reflect the results of the first to third phases of AI security testing and be published on the Financial Security Institute's website.
Does South Korea have an AI law that applies to banks?
Yes. The Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust (Act No. 20676), in force since 22 January 2026, applies to businesses that provide AI products or services, banks included. A bank that offers a service using generative or high-impact AI must notify users in advance, and AI used for judgments that significantly affect individuals' rights and obligations, such as loan screening, is listed as high-impact AI with duties on risk management, explanation, user protection, human oversight and documentation. The Ministry of Science and ICT enforces the Act.
Are the FSC's AI guidelines for the financial sector binding?
No. The FSC describes the revised guidelines of June 2026 as self-regulatory: each company decides how far to apply them according to its resources, the scope of its AI use and the risk of the service. The FSC adds that the use of high-impact AI or other AI regulated under the AI Basic Act must follow those legal requirements, which are binding.
What did Korean financial regulators do about Claude Mythos and frontier AI?
On 22 May 2026 the FSC eased its network separation rule, through one-year no-action letters, so that qualified large financial companies could test frontier AI models and SaaS tools for cyber defence and report their findings. On 2 July 2026 it waived sanctions for minor outages caused by AI security tests and urgent patches where service was restored quickly and consumers were protected, and issued a six-area response guide. A second phase opened to 75 companies on 3 September 2026.
Can Korean banks use ChatGPT or other cloud generative AI?
Only through an exception to the network separation rule. Since the FSC's August 2024 roadmap, financial companies can use internet-based generative AI by obtaining a regulatory sandbox exemption, with security measures checked by the FSS and the Financial Security Institute. The FSC's 2026 guidelines repeat that using commercial generative AI from the internal network requires designation as an innovative financial service, with enhanced security conditions.
Follow every move these regulators make
when one of these regulators moves, the next morning's brief says so · six sourced stories · 7 am PT · free
plus every regulation, bank and agent page update, the morning after · leave any morning