If AI Mis-Sells a Financial Product, Who Is Liable? [Bae, Kim & Lee’s Future Finance]

FSC Says AI Is a Support Tool, With Human Intervention Required

Financial Firms and Employees May Bear Greater Liability

Companies Need Rules on the Degree of Human Involvement

Hankyung Law & Biz’s “Law Street” column provides practical legal guidance for companies and individuals. Lawyers cover legal issues across tax, inheritance, labor, antitrust, M&A and finance, and also analyze major court rulings.

One of the early questions surrounding artificial intelligence was whether a driver or a manufacturer should be held responsible if a self-driving car caused an accident. The issue was never confined to autonomous vehicles. At its core was a broader debate over whether humans should be liable for machines or solutions that make decisions on their own. Self-driving cars simply became the most visible example.

As AI spreads, a similar debate is resurfacing. Questions are growing over who should be held responsible when AI wrongly recommends a financial product to a consumer or when a hiring process is conducted unfairly. In that context, the Financial Services Commission’s revised Guidelines on Artificial Intelligence in the Financial Sector, released this year, offer important implications.

FSC: AI is a support tool, and humans must be involved

The FSC guidelines set out standards for the administrative supervision of financial institutions, making them distinct from civil-law judgments. From a regulatory standpoint, however, their direction is clear.

The FSC identifies “support-tool status” as one of seven principles governing AI use. The guidelines state that financial companies should use AI as a support tool in their operations and build internal control systems so that final decision-making, and the responsibility that follows, remains with executives and employees. In particular, for high-risk AI, firms need to establish and operate standards that allow internal personnel to intervene in how the AI functions.

The principle means AI-generated output should be used as reference material, while human review and judgment continue throughout the entire work process. Under that approach, AI serves only to assist work performed by employees at a financial institution. If a financial accident occurs, that framework could lead to the institution bearing primary responsibility.

The guidelines also suggest something more. Because AI is to be used as a support tool in actual operations, responsibility for its use may rest more with the business unit that applied it than with the IT department that led its development.

Financial institutions and employees stand to shoulder more responsibility

From the perspective of financial institutions, that may seem unfair. Even after adopting AI, if the institution and its employees remain responsible, some may wonder whether it matters if they conduct advance testing before deployment.

Yet compliance efforts such as pre-deployment testing can still be fully reflected when regulators assess administrative liability. That makes preventive steps and post-incident controls crucial to avoiding accidents tied to AI services.

The FSC guidelines are not legally binding rules enacted under statute. The guidelines themselves therefore cannot determine civil liability. Unlike administrative responsibility, detailed discussion of civil damages arising from financial institutions’ use of AI has not yet developed sufficiently.

There is still little case law or academic consensus on who should bear responsibility when faulty AI output causes losses for a customer. It remains unclear whether liability rests with the financial institution, the employee who handled the work, or the outside vendor that supplied the AI model.

Still, individual financial companies are expected to reflect in internal rules a structure that assigns employees final responsibility for using and verifying AI output. That, in turn, raises the possibility that employees will be judged to have been meaningfully involved in the use of AI. The result could be a greater chance that both financial institutions and their employees face civil liability for damages when an accident occurs during AI use.

Rules on human intervention could help resolve liability questions

Financial services directly affect people to a significant degree. Given the current spread and sophistication of AI, the guidelines’ support-tool principle appears necessary in finance. It is more doubtful whether the same principle can be applied to other sectors.

The FSC’s support-tool principle is worth considering through the lens of responsibility for AI use. Liability for a company and its employees will vary depending on the degree of human involvement. Each company therefore needs to establish its own principles on that issue and build a compliance system capable of enforcing them. That could provide a simpler way to resolve liability questions arising from AI use.

<Hankyung Law & Biz Contributors> Yoon Joo-ho, attorney at Bae, Kim & Lee