What changed, and when.
- 13 Aug 2024Network-separation roadmap: generative AI with internet access allowed through a regulatory sandbox exemption, after security measures are in place.
- 18 Jun 2026The FSC announces updated AI guidelines for the financial sector.
- 22 Jun 2026The guidelines take effect: seven self-regulatory principles for all financial companies, fintechs included.
In the FSC’s words.
“At current stage, AI should be used as a means of assistance, which requires human supervision, and the ultimate responsibility for decision-making lies with the AI supervisor.”
On network separation, the 2024 roadmap allows generative AI that reaches the internet through the regulatory sandbox, “under the condition that financial companies prepare in advance sufficient security assurance measures”.
What it asks for.
The guidelines set seven principles. In the FSC’s English release they are:
- 1Governance: chief executives and executives pay keen attention to how AI is developed and used, with specific roles and responsibilities in place.
- 2Legitimacy: AI use complies with both financial and AI-related laws and regulations.
- 3Means of assistance: AI assists, under human supervision, and the ultimate responsibility for decisions lies with the person supervising it.
- 4Credibility: the data and models behind an AI agent are credible.
- 5Financial stability: risk to financial stability stays minimal at every stage, from design to learning.
- 6Good faith: AI is used in good faith, with the consumer’s interest first.
- 7Security: AI use follows security standards, with a way to keep examining and upgrading them.
“Individual companies are recommended to apply these guidelines on a discretionary basis considering the level of human and physical resources with which they are equipped, the scope of AI application, and the level of risk attached to the provision of service.”
The guidelines are self-regulatory. AI that counts as high-impact under Korea’s Framework Act on the Development of Artificial Intelligence follows that Act’s requirements. The FSS and the Financial Security Institute will each publish an AI risk management framework and an AI security guidebook for the sector.
What it means on a research desk.
A Korean research desk works inside network separation, where public AI tools on the work network need an exemption first. The guidelines add a second question to “can we use it”: who supervises it, and what record shows they did. Self-regulatory or not, that is the question an examiner or a client’s due diligence will ask.
- 1Every AI-written note gets a reviewer before anyone acts on it.
- 2There is a record of what was asked, what was answered, and who checked it.
- 3Confidential documents and positions stay inside the firm’s network.
How Core answers each requirement.
- Human supervision
- Every figure cites its source page, and a figure with no source is flagged, so the reviewer checks numbers in one click instead of retyping them.
- Responsibility, with a record
- Each Ask answer is logged with the model, where it ran, its sources and who reviewed it. The log is kept on the machine in your office.
- Network separation and security
- Core runs on hardware your firm owns, in your office, and your documents stay on it. What leaves depends on the setup, and we state it for each: on a Mac mini, questions go to a cloud model under your own agreement.
- Evidence for an examination
- An inspection pack in one download: the model card, controls, the use-case inventory and the answer log, framed on the SFC’s principles.
- Korean filings
- Core reads DART filings in full, in Korean, and cites the page each figure came from.
This is a summary of public documents, not legal advice. Check how they apply to your firm with your compliance team.
Questions.
When did the FSC AI guidelines for the financial sector take effect?
On 22 June 2026. The FSC announced them on 18 June 2026.
Are the FSC AI guidelines mandatory?
They are self-regulatory. The FSC recommends that each company apply them on a discretionary basis, considering its resources, the scope of its AI use and the risk of the service. AI that is high-impact under the Framework Act on the Development of Artificial Intelligence follows that Act.
What do the FSC guidelines say about human supervision?
The FSC’s release says that at the current stage AI should be used as a means of assistance, which requires human supervision, and that the ultimate responsibility for decision-making lies with the AI supervisor.
Can a Korean financial company use generative AI under network separation?
The FSC’s August 2024 roadmap allows generative AI that reaches the internet through a regulatory sandbox exemption, on condition that the company prepares sufficient security measures in advance. Whether your use needs an exemption is a question for your compliance team.
Who is responsible for an AI-assisted research decision in Korea?
Under the FSC guidelines, the person supervising the AI carries the ultimate responsibility for decision-making, not the AI. The Korean text puts it on the firm’s officers and employees.
