# Korea FSC Frontier AI Network Separation Easing: Network Separation Rules to be Eased in Financial Sector to Boost Innovation and Cybersecurity in AI Transformation

Source: https://www.bankingnewsai.com/ai-regulation/documents/korea-fsc-frontier-ai-network-separation-2026
Last updated: Nov 19, 2026

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. Companies that pass an expert screening receive a one-year no-action letter, must meet compensating security rules and must report their test findings to the government, which will use them to draw up cybersecurity guidelines for all financial sectors. The FSC also said it would consider lifting network separation completely, through the regulatory sandbox, for companies with advanced security and AI capabilities.

## At a glance

| Field | Value |
| --- | --- |
| Authority | [South Korea (FSC / FSS)](https://www.bankingnewsai.com/ai-regulation/south-korea) |
| Type | Guidance |
| Status | Final |
| Published | May 22, 2026 |
| Applies to | Financial companies under the Electronic Financial Transactions Act that must appoint a chief information security officer. The first phase was open to 49 companies with at least KRW10 trillion in assets and 1,000 or more regular staff; the second phase from September 2026 widened eligibility to 75, including nonbanks and electronic financial service providers. The relief is a temporary one-year no-action letter limited to AI and SaaS used for cyber defence. The FSC's English release is dated 25 May 2026 |
| Official text | https://www.fsc.go.kr/eng/pr010101/86977 |

## Key points

- Eligibility: 49 financial companies with at least KRW10 trillion in assets and 1,000 or more regular staff that come under the Electronic Financial Transactions Act and must appoint a CISO; an expert screening assesses each applicant's cybersecurity management and AI capabilities.
- Scope: the easing 'is only for the use of AI and SaaS programs for cybersecurity defense purposes to test vulnerabilities and use SaaS security tools', and selected firms must comply with security rules that make up for it.
- Reporting: after testing, firms report the characteristics of advanced AI cyber threats, the risks expected when such models are used in attacks and ways to defend, which the FSC will use for sector-wide cybersecurity guidelines.
- Phasing: about 10 companies in a first screening by June to July 2026, 10 to 20 in August to September and the rest in the fourth quarter; the Financial Security Institute would check AI vulnerabilities for up to 17 companies that did not apply, until July.
- Full lifting: for companies with advanced cybersecurity and AI capabilities, the FSC would consider removing network separation entirely through the financial regulatory sandbox, and extending that to more sectors in stages.
- Institutions: a technology advisory group of AI and security experts; the financial sector taskforce on advanced AI cyber threat set up in April 2026 with the FSC, FSS, Financial Security Institute and CISOs; a new research institution on AI security at the Financial Security Institute; and an AI cybersecurity support center.
- Support: detailed AI cybersecurity guidelines in June 2026 covering self-assessment, categorisation of computing infrastructure and patch prioritisation; reduced sanctions or exemption from liability for minor system errors caused by active security patching, on condition of prompt restoration and consumer protection; and funding for AI-driven security checks at small fintech firms.

## What changed for banks

Network separation had been relaxed case by case for generative AI since 2024; this measure created a separate track for frontier AI in cyber defence, using no-action letters. The follow-ups came quickly: on 30 June 2026 the FSC approved sanction relief for minor outages from AI security testing and patching and issued a response guide (published 2 July), on 3 September 2026 it opened a second phase to 75 eligible companies with 15 to be selected, and on 5 November 2026 it said it would seek to make AI-SaaS use for cyber defence permanent and to lift the rule completely.

## Use cases it governs

- [Cybersecurity](https://www.bankingnewsai.com/ai-regulation/by-use-case#cybersecurity)
- [Generative & agentic AI](https://www.bankingnewsai.com/ai-regulation/by-use-case#generative-agentic-ai)
- [Third-party & vendor AI](https://www.bankingnewsai.com/ai-regulation/by-use-case#third-party-vendors)
- [AI governance (general)](https://www.bankingnewsai.com/ai-regulation/by-use-case#governance-general)

## FAQ

### What did Korea's FSC do about Claude Mythos?

It treated frontier models such as Mythos as both a threat and a defensive tool. On 22 May 2026 it said financial companies must keep up regular cyber hygiene because they cannot be made entirely immune to these threats, and it eased network separation through one-year no-action letters so that qualified firms could use advanced AI models to find their own vulnerabilities and use SaaS security tools, reporting what they learn to the government.

### Which Korean financial companies can use frontier AI under the network separation easing?

In the first phase, the 49 companies under the Electronic Financial Transactions Act with at least KRW10 trillion in assets and 1,000 or more regular staff, about 10 of which were to be selected after expert screening. The second phase announced on 3 September 2026 lowered the bar to KRW2 trillion in assets and 300 staff, added electronic financial service providers with at least KRW2 trillion in annual transactions, and raised eligibility to 75 companies with 15 to be selected.

## Related documents

- [Korea FSC Frontier AI Security Response Guide](https://www.bankingnewsai.com/ai-regulation/documents/korea-fsc-frontier-ai-response-guide-2026) — Financial Sector Response Guide to Frontier AI Security Threats (프런티어 AI 보안위협 금융분야 대응요령) (Jul 2, 2026)
- [Korea FSC AI Guidelines for the Financial Sector](https://www.bankingnewsai.com/ai-regulation/documents/korea-fsc-ai-guidelines-2026) — AI Guidelines in the Financial Sector (금융분야 인공지능 가이드라인) (Jun 18, 2026)
- [Korea FSS AI Risk Management Framework](https://www.bankingnewsai.com/ai-regulation/documents/korea-fss-ai-rmf-2026) — AI Risk Management Framework in the Financial Sector (금융분야 AI 위험관리 프레임워크, AI RMF) (Jan 15, 2026)
- [Korea AI Basic Act](https://www.bankingnewsai.com/ai-regulation/documents/korea-ai-basic-act-2025) — Framework Act on the Development of Artificial Intelligence and the Creation of a Foundation for Trust (인공지능 발전과 신뢰 기반 조성 등에 관한 기본법) (Jan 21, 2025)
- [Korea FSC Network Separation Roadmap](https://www.bankingnewsai.com/ai-regulation/documents/korea-fsc-network-separation-roadmap-2024) — Roadmap to Improve Network Separation in the Financial Industry (금융분야 망분리 개선 로드맵) (Aug 13, 2024)
- [Korea FSC AI Credit Scoring and AI Security Guideline](https://www.bankingnewsai.com/ai-regulation/documents/korea-fsc-ai-credit-scoring-security-2023) — Verification System for AI-driven Credit Scoring Models and Security Guideline for the Use of AI in the Financial Sector (Apr 17, 2023)
- [Korea FSC AI Guideline 2021](https://www.bankingnewsai.com/ai-regulation/documents/korea-fsc-ai-guideline-2021) — Guideline on the Use of Artificial Intelligence in Financial Services (금융분야 인공지능(AI) 가이드라인) (Jul 8, 2021)
- [BNM AI Feedback Statement](https://www.bankingnewsai.com/ai-regulation/documents/bnm-ai-feedback-statement-2026) — Feedback Statement: Artificial Intelligence (AI) in the Malaysian Financial Sector (Nov 19, 2026)

Last reviewed Nov 19, 2026. Cite the official text (https://www.fsc.go.kr/eng/pr010101/86977) for the rule and this page for the summary and dates.

---

Canonical page: https://www.bankingnewsai.com/ai-regulation/documents/korea-fsc-frontier-ai-network-separation-2026
Part of [BankingNewsAI](https://www.bankingnewsai.com/) — a free daily brief on AI in banking, an AI regulation tracker (60 authorities, 360 documents) and AI-strategy profiles of the 120 largest US banks. Markdown versions of every reference page: append `.md` to the page URL; index at https://www.bankingnewsai.com/llms.txt.
