90 victims, 8.2 million records: what South Korea's first AI hacking case reveals

South Korea's Seoul Metropolitan Police Agency cyber investigation division announced on October 26 that it had apprehended seven suspects, including a 32-year-old alleged ringleader identified only as "A," over a hacking and extortion campaign that used generative AI to break into about 90 domestic organizations. Police referred the detained ringleader to prosecutors on October 23 on charges including violating the Information and Communications Network Act and extortion or attempted extortion. The remaining six suspects are under investigation without detention.

The group is accused of stealing more than 8.2 million personal data records and internal documents between September 2025 and October 2026, a span of about 13 months. Victims included cram schools, hospitals, golf courses, universities, and affiliates of major South Korean conglomerates. Police say 60 of the victim organizations were attacked with the help of AI, and that an "AI autonomous penetration testing system" previously used in financial-sector data breach incidents was also deployed.

The extortion component was systematic. The group sent threatening emails to 38 victim companies demanding a total of 360 million won, roughly $270,000. Police are investigating whether the stolen data was leaked a second time.

What makes this case notable is not the scale alone but the profile of the alleged attacker. The ringleader reportedly had no formal IT training and learned hacking techniques from blogs. Police describe the case as South Korea's first arrest in a large-scale hacking and personal data theft case involving the misuse of generative AI, and say the suspect sharply expanded the scale of the hacking in a short period after beginning to make full use of AI.

That detail matters for defenders. If a self-taught operator can allegedly compromise roughly 90 organizations and exfiltrate 8.2 million records in about 13 months, the technical barrier to industrial-scale cybercrime is lower than many security teams assume. Smaller organizations with weaker security postures — the hospitals, academies, and local businesses that make up much of the victim list — face the sharpest risk, because they often lack the monitoring and incident response capacity that larger enterprises maintain.

The case also raises questions that remain unanswered. How exactly generative AI enabled the intrusions — whether for reconnaissance, vulnerability discovery, exploit generation, or automation — has not been detailed beyond the reference to an AI autonomous penetration testing system.