Cheong Wa Dae AI chief warns finance, defense most vulnerable as AI-driven hacking becomes routine

Amid a string of AI-assisted cyberattacks targeting financial institutions and major Seoul churches, Cheong Wa Dae Senior Presidential Secretary for AI Future Planning Lee Hae-min warned Friday that hacking has become an everyday occurrence, with the finance and defense sectors facing the gravest risks.

Speaking on SBS Radio's "Kim Tae-hyun's Politics Show," Lee said more than 30 cases of personal data breaches or hacking attempts are occurring daily. "The frequency has gone up," he said.

"I thought the most powerful AI could provide the most powerful security, but looking at the current situation in South Korea, I concluded that finance and defense are the two most dangerous areas," Lee said. The concern had prompted him, during his time as a lawmaker, to transfer to the National Assembly's Political Affairs Committee to keep a closer eye on the financial sector. "Seeing the financial sector getting hit now, I thought — this is serious, it came a little faster than I expected," he said.

On the dynamics of AI-based attack and defense, Lee said AI can serve as both a spear and a shield. "The spear and the shield are locked together. Ultimately, AI attacks will have to be stopped by AI," he said. He also stressed the need to shift to a standing alert posture before an incident occurs, rather than responding only after the fact.

Lee was particularly critical of the existing government governance structure for cybersecurity. "In this new era of AI hacking, the old setup — where the Ministry of Interior and Safety handles administrative networks, the Personal Information Protection Commission handles personal data, and KISA and the Ministry of Science and ICT handle the private sector — no longer fits," he said. He said the overall structure needs to be reorganized along the lines of an emergency disaster-response system, and that work is already underway.

On allegations surrounding the use of Chinese open-weight AI models in the attacks, Lee said the models are available worldwide and that determining who used them and where falls within the scope of a criminal investigation. "The investigation side and the side focused on maintaining security need to be kept somewhat separate," he said. He added that his office is concentrating on how to defend against hacking that exploits open-weight, open-source or publicly known AI models.

Lee also took a critical view of the financial sector's existing network-separation policy. "We implemented network separation in the name of security. At first it feels like an incredibly solid fortress, but the inside ends up becoming a sterile room — which actually makes it more dangerous," he said. He added that the entire system needs to be redesigned from the ground up to incorporate both offensive and defensive concepts, and that this effort is also currently in progress.