Bank security has long assumed that serious attacks take serious teams. A string of breaches in South Korea suggests that assumption is getting old.

On October 7, cybersecurity firm CrowdStrike published findings on a campaign against South Korean financial firms that ran from late September into early October. Its conclusion: one person, using AI agents, carried out multiple intrusions in a short span of time.

What happened

At least nine South Korean banks have disclosed or been reported as targets since late September, Reuters reported. Shinhan Bank said personal information of about 25,000 customers was compromised. KB Kookmin reported 119 affected customers, and Hana Bank 89, according to The Register, citing The Korea Times.

The entry points were unglamorous. At one bank, the attacker got in through a loan-status lookup service used by brokers. At another, it was an employee mobile work app.

The tool at the center was ARTEX, an open-source "agentic" penetration-testing program published on GitHub this year by a Chinese developer. Penetration testing is normally done by security teams probing their own systems for weaknesses. ARTEX connects to outside AI models that plan and run those probes with little human help.

CrowdStrike said the attacker mainly used DeepSeek, a Chinese AI model, along with other models. The attacker also left files exposed on their own servers, including session logs from Anthropic's Claude Code. Reuters reported the attacker asked Claude where stolen Korean data is typically sold. Anthropic did not respond to Reuters' request for comment.

Those exposed logs also gave investigators clues to the attacker's identity. CrowdStrike assessed, with moderate confidence, that the actor is likely a Chinese speaker motivated by money. China's foreign ministry said it opposes hacking.

The fallout

South Korean police opened an investigation, and President Lee Jae Myung called for a strong response. Lawmakers voted to summon the heads of five major banks to a parliamentary audit on October 19.

CrowdStrike's Adam Meyers, who leads its counter-adversary work, described it as a single human using AI agents to run widespread attacks. CrowdStrike expects other attackers to keep experimenting with AI to move faster.

Why it matters

The story isn't that AI invented a new kind of attack. It's that AI let one person do the work of a team, testing many doors quickly until one opened. That changes the math for every bank, especially smaller ones with thin security staffs.

It also turns a defensive tool into an offensive one. Agentic testing software is built to help companies find their own weak spots before criminals do. The same software, pointed at someone else's systems, does the criminal's homework instead. ARTEX's own GitHub page says it is meant for personal learning and should not be used against live systems. That warning did not stop anyone.

U.S. regulators are already paying attention. In June, Reuters reported that the OCC and Federal Reserve had started asking banks during routine exams to map how they use AI in high-risk areas. Cyber insurers, meanwhile, are reviewing whether their policies cover losses caused by autonomous AI systems.

What you can do

You can't control your bank's servers. You can limit the damage when a breach exposes your information.

Freeze your credit. A credit freeze has been free by federal law since 2018 at all three major credit bureaus. It blocks most new credit accounts from being opened in your name.

Turn on transaction alerts in your bank and card apps so unfamiliar charges surface right away.

Use strong, unique passwords and passkeys where your bank offers them, so one leaked credential doesn't open everything.

Expect follow-up scams. After big breaches, criminals often pose as the bank. Never act on an urgent call or text. Hang up and call the number on the back of your card.

What to watch

Watch what Korean regulators demand of their banks after the October 19 audit. Watch also for U.S. supervisors to address AI-driven attacks in formal guidance, not just exam questions.

Sources: CrowdStrike (Oct. 7) · Reuters via Insurance Journal (Oct. 8) · The Register (Oct. 8)

Beyond the Bank is for informational purposes only and is not financial, legal, or investment advice.

Get Beyond the Bank in your inbox every Monday. A free, plain-English briefing on AI, banking alternatives and the new rails under your money. Read the latest issue or subscribe free.