Artificial intelligence is lowering the barrier to financial-sector cyberattacks while also becoming a tool for finding weaknesses, according to discussions at FISCON 2026 in Seoul. Speakers at the event, hosted by the Financial Security Institute, said attackers need to identify only one exploitable weakness, while defenders must also assess the effects of a repair across connected systems.
An analysis presented at the event found that the hacker interface of the Chinese-linked open AI tool Artex showed 255 targets during August and September, including eight targets in South Korea. More than half of the Korean targets were in the financial sector, although the report did not identify confirmed breaches for those cases.
The Financial Security Institute also described an exercise recreating the theft of about $3.1 million in a day from 431 virtual-asset wallets. In that test, AI found the password in as little as two seconds and as long as seven or eight minutes. The result demonstrates the speed claimed in the exercise, but does not establish that the same method caused a real-world financial incident.
Defensive testing produced a more limited but concrete result. Using the generative AI system Claude across 16 financial companies, researchers found that AI contributed to identifying 34 of 94 vulnerabilities. Its strongest role was in repetitive work such as API checks, where it found weaknesses that people had missed. The institute’s president apologized over a separate internal hacking incident and said all systems would undergo a full inspection, followed by measures to prevent a recurrence.
