An unidentified hacker launched targeted attacks against multiple financial institutions in South Korea from late September to early October, stealing data in the process and using LLMs as well as the open-source AI penetration-testing tool ARTEX.

What’s truly noteworthy isn’t just that “the hacker used AI,” but that AI is beginning to enter the attack chain.

Previously, penetration work relied more on manual analysis, scripts, and fixed procedures. Now, models can help with understanding targets, orchestrating steps, and adjusting actions based on feedback.

AI may not create new vulnerabilities, but it can lower the threshold for using existing tools—making tasks that previously required coordination by professional teams easier for a small number of people to carry out.

And open-source tools further amplify this shift. The tools themselves may not be malicious. However, when open-source software lowers technical barriers, and LLMs lower operational barriers as well, offensive capabilities can spread beyond a small group of experts.

So what these reports should genuinely raise concern about isn’t the gimmick of “AI hackers,” but the fact that the cost structures of both attackers and defenders are changing.

While AI lowers the productivity barrier, it may also lower the cost of attacks.