South Korean Banks Likely Hacked by China-Based Actor Using AI Agent, CrowdStrike Says
A suspected China-based hacker used an artificial-intelligence agent and multiple large language models to target South Korean financial institutions in a campaign that exposed a growing risk for banks: cyberattacks that can be scaled by a single person using autonomous AI tools.
Cybersecurity company CrowdStrike said the activity targeted multiple South Korean financial organizations between late September and early October 2026. The company assessed with moderate confidence that the threat actor was likely Chinese-speaking and financially motivated, but it did not attribute the campaign to a named hacking group.
The findings add a new dimension to a series of recent data breaches involving South Korean banks and come as governments and financial institutions worldwide try to understand how quickly AI agents can change the economics of cybercrime.
What CrowdStrike found
CrowdStrike said the attacker used ARTEX, an open-source agentic penetration-testing tool developed in China, alongside large language models.
Unlike a conventional AI chatbot that simply responds to individual prompts, an agentic AI system can perform a sequence of tasks with greater autonomy. In cybersecurity, that can include identifying potential weaknesses, gathering information and helping execute technical steps.
CrowdStrike's investigation uncovered attacker-controlled open directories containing Claude Code session histories, ARTEX configuration files and Claude memory files. These materials gave researchers a detailed view of the tools and workflow used during the campaign.
The activity involved a two-server setup, according to CrowdStrike. One Hong Kong-based server appeared to serve as the attacker's primary infrastructure, while another server hosted an ARTEX instance that was likely used in the attacks against South Korean financial organizations.
Was the attacker definitely based in China?
Not yet.
This distinction is important.
CrowdStrike has not publicly identified the attacker by name and has not said that the Chinese government was behind the campaign.
Instead, the cybersecurity firm assessed with moderate confidence that the actor was likely a Chinese speaker and financially motivated. The assessment was based partly on the use of the Chinese-developed ARTEX tool and Chinese-language prompts observed in the AI sessions.
CrowdStrike also found information in one Claude Code session that appeared to identify a possible attacker as a 26-year-old with educational ties to Guangdong, China. Reuters reported that the individual may have been based in the city of Maoming, but the person's identity and connection to the attacks remain unconfirmed.
That means the appropriate description at this stage is a likely China-based or Chinese-speaking individual, rather than a confirmed Chinese state-sponsored operation.
China's Foreign Ministry has also said it opposes and combats hacking activities.
How AI was used in the attacks
The campaign is significant because the attacker appears to have combined several AI systems rather than relying on a single model.
CrowdStrike said the ARTEX instance primarily used DeepSeek v4.1-flash as its LLM backend. The attacker also used GLM-5.3 and Grok 4.6 during Claude Code sessions.
This illustrates how AI tools can be combined into a broader workflow.
ARTEX itself is not a standalone large language model. It is an agentic penetration-testing tool capable of connecting to external AI models. Its stated purpose is security testing, learning and technical research, but the South Korean campaign demonstrates how tools created for legitimate cybersecurity work can potentially be repurposed for malicious activity
That dual-use problem is becoming increasingly important for cybersecurity companies.
South Korean banks were already dealing with data breaches
The CrowdStrike findings come after a series of cyber incidents affecting South Korean financial institutions.
Shinhan Bank said personal information belonging to about 25,000 customers had been compromised. KB Kookmin Bank said personal information involving 119 customers was leaked. Other financial institutions, including Hana Bank, were also reported to have experienced breaches or attacks.
South Korean authorities launched investigations after the incidents, while President Lee Jae Myung called for a stronger response after officials said AI appeared to have been involved in the attacks.
The affected systems were not necessarily the banks' core banking systems. Reported targets included a loan inquiry service used by financial brokers and an employee mobile work-support system.
That distinction matters because it shows that attackers do not necessarily need to penetrate a bank's central infrastructure to obtain valuable information.
The bigger threat: one attacker can do more
The most important implication for financial institutions may be the scale AI can provide to an individual attacker.
Adam Meyers, CrowdStrike's senior vice president of counter adversary operations, described the campaign as an example of a human attacker using AI agents to conduct attacks at scale. The technology can potentially allow one person to target many organizations in a much shorter period of time.
Traditionally, sophisticated cyberattacks often required teams with specialized skills to conduct reconnaissance, develop tools, identify vulnerabilities and move through compromised systems.
Agentic AI could reduce some of that workload.
That does not mean AI automatically turns an inexperienced person into an elite hacker. Real-world attacks still face security controls, authentication barriers, monitoring systems and other technical obstacles.
But AI can potentially make repetitive and time-consuming parts of an attack faster.
For banks, that changes the threat model.
Why this matters to cybersecurity companies
The incident could strengthen demand for technologies designed to detect unusual behavior, protect identities and monitor AI-enabled attack techniques.
Cybersecurity companies such as CrowdStrike are already positioning AI as both a defensive tool and a new dimension of the threat landscape. CrowdStrike's 2026 financial-services threat research said adversaries were increasingly using AI to shorten the time between gaining access and causing impact.
The South Korean incidents provide a real-world example of why financial institutions may need to increase spending on AI-assisted security, threat intelligence, identity protection and continuous monitoring.
However, investors should not assume that every AI-related cyberattack automatically translates into higher revenue for cybersecurity companies. Spending decisions depend on bank budgets, regulatory requirements, competitive dynamics and whether vendors can demonstrate measurable protection.
What investors should watch
The investigation could have implications beyond South Korea.
Investors following cybersecurity and financial technology should watch several developments:
Further attribution: Whether South Korean authorities or cybersecurity researchers identify the attacker or a broader network.
Extent of stolen data: The full scale of compromised customer and corporate information remains unclear.
AI-agent attacks: Whether similar campaigns emerge against banks in other countries.
Bank cybersecurity spending: Financial institutions may increase investment in AI-driven detection and response.
Regulatory action: Authorities could introduce stricter requirements for financial institutions' cybersecurity defenses.
Cyber-insurance risks: Insurers may reassess how policies treat attacks involving autonomous AI systems.
AI safety controls: Developers of AI and security tools may face greater pressure to prevent their systems from being misused.
The cyber-insurance issue is particularly interesting. Reuters reported that the emergence of autonomous AI systems is forcing insurers to consider questions around what constitutes an attacker and who bears liability when AI-generated actions contribute to a breach.
AI is changing both sides of the cybersecurity battle
The South Korean banking incidents highlight a broader shift in cybersecurity.
AI is not inherently a hacking technology. The same tools can help defenders analyze large amounts of data, identify suspicious behavior and automate security testing. But attackers can also use AI to accelerate reconnaissance, coding and operational tasks.
That creates an arms race in which banks and cybersecurity providers may increasingly need AI-powered defenses to respond to AI-assisted attacks.
For financial institutions, the objective is therefore not simply to block a particular tool such as ARTEX. It is to detect the behavior that emerges when multiple AI systems, compromised infrastructure and traditional hacking techniques are combined.
Bottom line
CrowdStrike says a likely Chinese-speaking, financially motivated actor used ARTEX and multiple AI models in attacks against South Korean financial organizations between late September and early October. The company has not attributed the campaign to a named adversary, and the identity of the suspected individual remains unconfirmed.
The significance extends beyond the affected Korean banks. The case demonstrates how agentic AI could allow a single attacker to automate parts of a sophisticated cyber campaign and potentially target multiple organizations more efficiently.
For investors, the key issue is whether these incidents become isolated cases or part of a broader trend that pushes banks, governments and corporations to accelerate cybersecurity spending. The answer could shape the next phase of growth for the global cybersecurity industry.
Follow the blog for more updates and analysis on cybersecurity, AI, technology companies, banks and global markets.
This article is for informational and educational purposes only and should not be considered investment advice.