IncidentAI-enabled attacks
South Korea: Shinhan, KB Kookmin and Hana breached through internet-facing helper systems, with traces of an AI attack agent
Between September 30 and October 1, 2026 outsiders reached internet-facing services at three of South Korea's largest banks: a mobile inquiry service for Shinhan Bank's loan agents (names, phone numbers, annual income and loan limits of about 25,000 customers, and 66 resident registration numbers), KB Kookmin Bank's employee mobile support system (119 customers) and Hana Bank's sales support system (89 customers); BNK Financial Group reported 11 records of outsourced staff. Yonhap reported that the attackers probably used sophisticated AI agents to probe for weaknesses, and that a server used in the attacks carried a page title associated with ARTEX AI, an open-source agentic penetration-testing system. Neither the banks nor the authorities have confirmed the tool was used. The Financial Services Commission held an emergency meeting on October 2 and ordered every financial company to review its externally accessible systems. On October 6 President Lee Jae Myung told a cabinet meeting that "In some hacking incidents, signs have emerged of AI being used, causing considerable public concern"; police opened a full-scale investigation, Yonhap reported that Woori Bank had been breached too, and the Financial Supervisory Service and the Financial Security Institute shared 28 attacker IP addresses and some country information with the sector. Authorities have not said which AI tools were used. By the regulators' count reported on October 4 and 5, seven firms were breached: the three banks, BNK Busan Bank, Yegaram Savings Bank (about 40,000 people), Welcome Savings Bank (about 2,200) and Hyundai Capital (146), with KB Kookmin's count at 153. The same attacker IP address showed up at every bank breached and different ones at the savings banks; the targets were inquiry systems without identity checks, staff services without access controls and websites with unpatched flaws, and about 500 financial companies were told to finish security reviews by October 8. The Korea Herald's October 5 tally lists Woori Bank and NH NongHyup Bank among those that blocked the attempts with no data lost. On October 6 the Korean National Police Agency put 28 investigators from its cyberterrorism unit on the case; the Korea Herald put Shinhan's count at 25,729 customers and the exposure at about 66,000 people and 2,200 corporate records, and the Financial Services Commission told AFP the total is above 68,000. A government official told AFP it was "highly likely" that Artex AI, an open-source testing tool, was used, adding that this does not show who the attackers were. On October 7 CrowdStrike reported open directories on attacker servers, one running an ARTEX instance and another holding Claude Code session histories, Claude memory files and ARTEX configuration files, from a campaign active from late September to early October that resulted in exfiltrated data. ARTEX ran mainly on DeepSeek v4.1-flash, with GLM-5.3 and Grok 4.6 in other sessions, and the actor also asked Claude where stolen Korean breach data is sold. CrowdStrike named no adversary and assessed with moderate confidence that the actor is financially motivated. The Hacker News reports that the ARTEX developer said the misuse went against the tool's purpose and is moving the project to closed source with no further updates. On October 8 Reuters reported that CrowdStrike's analysis of the Claude Code sessions points, with moderate confidence, to a Chinese-speaking, financially motivated actor who may be a 26-year-old in Maoming, Guangdong: in one session the person asked Claude to write a security researcher's résumé with a Telegram account, age, education and location, and in another asked where Korean breach data is sold. A man who answered the phone number CrowdStrike gave said he knew nothing of it; Anthropic, the Korean police and China's foreign ministry did not respond to Reuters. By October 10 the Korea JoongAng Daily reported that the suspect denied involvement and said his identity had been stolen, that Logpresso had found him active in a Chinese DDoS-for-hire group, that the ARTEX developer had taken the tool closed source citing misuse and its GitHub page was down, that at least nine Korean financial institutions had been hit or targeted since late September, and that China's foreign ministry said it was unfamiliar with the case. On October 22 the Korea Herald reported that regulators have confirmed hacking incidents at 16 financial companies exposing about 100,000 people's data, not all tied to the same campaign, and quoted CrowdStrike's Adam Meyers that AI gave a relatively unskilled attacker "All of it"; several institutions were entered through systems used by staff or loan brokers rather than core banking platforms, and Woori Bank credited extra employee authentication for losing no data.
Replay the attack step by step in the Learning Center
How it happened
- The attackers reached services built for loan agents, staff and sales support that were reachable from the internet.
- Yonhap reported they probably used AI agents to probe for vulnerabilities; ARTEX AI's agents automate information gathering, vulnerability discovery, attack-path planning, tool execution and verification.
- Hana detected its intrusion the day after Shinhan's, having checked its systems when it learned of attempts at another institution.
Impact
Customer records, including income and loan limits for about 25,000 Shinhan customers and resident registration numbers in all three banks; regulators inspected the banks and ordered sector-wide checks.
The fix
The banks blocked the attackers' IP addresses or the affected servers, and Shinhan suspended its service. The FSC told financial companies to inspect all externally accessible systems, reduce unnecessary information exposure, check for missing or inadequate authentication and share threat information such as IP addresses and attack methods quickly.
Lessons
- Helper services built for agents, staff and partners face the internet too; they need the same authentication and monitoring as the main app.
- An automated attack agent tries many organizations at once: when one reports an intrusion, the others should check the same kinds of systems that day, as Hana did.
- Traces of a tool on an attack server point to the tool, not to who ran it.
Sources
- BleepingComputer: South Korea probes bank breaches amid suspected AI-powered attacks
- The Korea Herald: Hana, BNK report data breaches as hacking hits Korean banks
- Bloomberg via Insurance Journal: AI tools suspected in Korea's Shinhan Bank hack, Yonhap says
- Reuters, via WKZO: South Korea's Lee says AI appears to have been used in bank hacks
- Seoul Economic Daily: same IP used in hacks on multiple Korean financial firms
- The Korea Herald: Korean banks on high alert after wave of cyberattacks
- Financial Services Commission (Korean): emergency meeting on financial-sector cyber threats
- The Korea Herald: Police launch major probe as suspected AI hacks sweep through banks
- France24 (AFP): South Korea warns of possible AI use in banking hacks
- CrowdStrike: Unknown threat actor uses ARTEX to target South Korean finance
- The Hacker News: ARTEX AI pentesting tool used in data theft attacks on South Korean financial firms
- Reuters via Business Day: CrowdStrike says China-based suspect used AI tools in South Korean bank hacks
- Korea JoongAng Daily: Chinese AI tool linked to Korean bank hacks halts public updates as suspect denies involvement
- The Korea Herald: AI gave the hacker the tools, weak security opened the door, says CrowdStrike