ARTEX Developer Shuts Down Public AI Agent After CrowdStrike Reports South Korean Bank Cyberattacks

The developer of ARTEX, a Chinese-developed artificial intelligence agent identified by cybersecurity investigators as a tool used in a recent cyberattack campaign against South Korean banks, has withdrawn the software from public development and announced plans to convert it into a closed-source project.

The developer, who uses the GitHub handle "Autumn-27", announced the decision on Thursday, saying the project would no longer receive public updates, new releases, or maintenance support following reports of its alleged misuse.

"Given the misuse of the tool, the ARTEX project will no longer be updated and will be converted to closed source. No further versions will be released to the public nor will maintenance support be provided," the developer said on GitHub.

The announcement followed a report by US cybersecurity firm CrowdStrike, which said a suspected attacker targeting South Korean banks had used ARTEX alongside Anthropic's Claude Code. CrowdStrike assessed that the individual was likely a 26-year-old based in China, although it did not attribute the campaign to a named adversary.

The developer did not explicitly address the attacks on South Korean banks but said ARTEX was originally created to help enterprises and other organizations test their security risks and strengthen their cybersecurity capabilities. They also opposed any illegal use of the software, stating that they bore no responsibility for conduct that violated laws and regulations.

ARTEX's GitHub page had been taken down by Thursday, according to Reuters' checks, removing the publicly accessible project from the platform where it had been released earlier this year.

The withdrawal marks a significant development in the debate over the risks posed by AI-enabled cybersecurity tools. Software designed to help organizations identify vulnerabilities can also be repurposed by malicious actors, raising questions about how developers should respond when legitimate security tools become associated with criminal activity.

ARTEX is an open-source AI agent designed to automate penetration testing, a cybersecurity practice in which authorized testers probe computer systems for weaknesses that could be exploited by attackers.

Unlike a standalone large language model, ARTEX connects to external AI models, including ChatGPT, Claude, and DeepSeek, to assist with security testing and vulnerability assessment. Its ability to draw on these models allows it to combine automated processes with the coding and analytical capabilities of large language models.

CrowdStrike said the suspected attacker used ARTEX and Anthropic's Claude Code as part of a campaign aimed at stealing personal information from South Korean bank customers.

At least nine South Korean banks have disclosed or were reported by local media to have been targeted by cyberattacks since late September. The incidents prompted South Korean police to open an investigation this week, while President Lee Jae Myung called for robust response measures.

Shinhan Bank previously reported that personal information belonging to about 25,000 customers had been compromised, while KB Kookmin Bank said information belonging to 119 customers had been leaked.

CrowdStrike's findings highlighted how AI agents can potentially allow a single operator to conduct activities at greater speed and scale by automating parts of the technical work involved in cyberattacks. The use of ARTEX in the reported campaign has also drawn attention to the dual-use nature of AI tools that can support both legitimate security testing and malicious operations.

However, the available findings do not establish that ARTEX independently initiated the attacks or that the software was responsible for every stage of the campaign. The reported use of the tool instead points to a human operator leveraging AI-enabled capabilities as part of a broader operation.

AI agents are becoming more capable of carrying out multi-step technical tasks. Organizations developing these systems face the challenge of making them useful for defensive security work without making it easier for attackers to automate harmful activities.

Closing The Project Raises Questions About Open-Source Security

The developer's decision to stop public releases removes one avenue through which users could access and update ARTEX, but it does not necessarily eliminate copies already obtained or prevent similar tools from being developed elsewhere.

Converting the project to closed source also changes how the software can be examined and maintained by the wider security community. Open-source tools can be inspected, tested, and improved by independent researchers, but their accessibility also means that potentially dangerous capabilities can be obtained without the developer's direct involvement.

Closing a project may reduce access to future versions and official support, but it cannot by itself resolve the broader challenge of controlling dual-use AI technology. Existing copies may continue circulating, while comparable capabilities can be built using other AI models and automation frameworks.

The episode therefore illustrates a difficult trade-off for developers of AI-enabled security tools: keeping software publicly available can support legitimate research and defensive testing, while withdrawing it may limit some forms of access but also restrict legitimate users' ability to inspect and maintain the project.

The developer's announcement did not specify whether the closed-source version would remain available to selected users, how access would be controlled, or whether additional safeguards would be introduced. Those details will determine how substantially the decision changes the tool's future availability.

China's government has also distanced itself from the specific case. Foreign Ministry spokesperson Mao Ning said on Thursday that the ministry was not familiar with the matter and reiterated that China consistently opposes and combats hacking activities.

However, CrowdStrike's assessment that the suspected attacker was based in China remains an attribution judgment rather than a definitive identification of the individual responsible. The cybersecurity firm said its investigation uncovered details it believed were linked to a 26-year-old in Guangdong province, but it did not name a confirmed adversary.

The ARTEX developer's decision is consequently best understood as a response to reported misuse rather than proof of responsibility for the South Korean attacks.

More broadly, the incident indicates that the spread of AI agents is complicating the distinction between cybersecurity research and malicious activity. As tools become more capable of automating technical tasks, developers, businesses and regulators face growing pressure to determine how those capabilities should be distributed, monitored and restricted.

ARTEX's removal from public GitHub access may limit the availability of future official releases, but the underlying security challenge remains. Defenders must boost preparation for attackers who can combine conventional hacking techniques with AI-powered automation, while developers must consider how their tools could be used beyond their intended purposes.