The task force for building “KPS AI AEGIS” cheers after its first meeting on the 2nd of this month. Photo courtesy of KEPCO KPS
NAJU — KEPCO KPS, which has introduced artificial intelligence into its internal audit work, is now building a shield to control the risks that come with it.
The company said on the 8th that it is broadening the use of its audit-focused AI agent “Gamsahae” while starting work on a defense system to reduce potential risks such as personal data leaks and data security breaches.
KEPCO KPS has operated Gamsahae, an AI agent tailored to audit work, since May this year. Unlike general generative AI services, it is an internal support tool built around the practical needs of auditors.
Its main functions include helping draft audit reports, reviewing sentences and grammar, analyzing past disciplinary cases, generating reports on disciplinary levels, answering audit-related questions, checking the adequacy of construction designs and reviewing safety management costs. It also detects irregular signs and recurring patterns through data analysis, flagging risk signals and analytical findings to auditors.
The company said the tool has improved both the speed and accuracy of audit work and has contributed to raising the integrity of the organization.
But the wider use of AI has also raised new challenges. Ethical and technical risks that can arise during actual audits — errors, exposure of personal data, the learning of sensitive information, data security and the question of who is accountable for AI judgments — now require systematic management.
In response, KEPCO KPS formed a task force to build “KPS AI AEGIS” and held its first meeting on the 2nd of this month. Aegis refers to the shield that protects the gods in ancient Greek mythology, and the name reflects the company’s intent to protect the organization from risks created by AI. The task force is led by the audit office and includes relevant departments from across the company. Its goals are to establish standards for using AI and to design the operating model best suited to audit work.
KEPCO KPS plans to adopt an AI audit framework that puts the safety and reliability of the technology first, in line with South Korea’s national AI ethics principles. The principle of keeping a human in the loop was also a key agenda item at the meeting.
“Because the AI agent we introduced into audit work handles sensitive information, it requires a high level of ethics and thorough risk management,” said Lee Sung-kyu, standing auditor at KEPCO KPS. “We have to consider not only the fairness and reliability of what the AI produces but also the security of key information. We will move beyond being simple AI users to become ‘AI actors’ who use AI with critical thinking.”
