1. What did GMO Cybersecurity find?
The ARTEX AI tool has been tied to intrusions against Japanese businesses. On October 19, 2026, Tokyo-based GMO Cybersecurity by Ierae said it had confirmed traces of ARTEX, an AI-driven penetration-testing program developed in China, on servers used to attack Japanese companies. The infrastructure it examined was aimed at about 20 firms.
Penetration-testing software is designed for security professionals, who use it to probe their own systems for weaknesses before real attackers do. The finding is notable because it shows such a program being used in the opposite way, as part of the machinery behind actual break-ins.
The disclosure came from a Japanese security firm's own analysis of attacker servers, and it adds a Japanese chapter to a story that had until now centered on South Korea.
Researchers reach such conclusions by examining the servers that attackers use to stage and run their operations. Finding a tool's traces on that infrastructure indicates which software was used to carry out the attack, and the pattern of targets reveals who the operators were interested in.
2. How many companies were targeted, and which ones?
GMO's central finding concerns a single attack infrastructure that targeted roughly 20 Japanese companies. Nine of those companies operate in the same sector, although the sector was not identified in the account of the research. That concentration is one of the details that security teams will study most closely.
Eight of the targeted firms reported data breaches between September and October 2026. The timing lines up with the activity GMO observed on the infrastructure it examined, which is why the researchers treat the two as related. Matching dates do not by themselves prove that every breach was caused by the same operation, but they point in that direction.
The researchers also identified six further AI-based attack infrastructures that were aimed at Japan. Those servers were directed at organizations in retail, education, finance and media, which means the activity is not confined to one industry or one type of victim.
For the companies involved, a data breach carries consequences that go beyond the technical incident. Customers, partners and regulators expect prompt disclosure, and firms must work out what information was taken and how to contain further access. Eight disclosures within roughly two months is a notable cluster for a country's corporate sector.
3. What is the ARTEX AI tool?
ARTEX is an agentic penetration-testing tool, meaning that artificial intelligence drives the steps of a security test with limited human direction. Instead of an operator running each scan and exploit by hand, the software can chain together reconnaissance, probing and exploitation on its own, which speeds up work that would otherwise take a skilled person considerably longer.
That capability is valuable to defenders and attackers alike. A tool that can survey a network, identify weak points and attempt to use them automatically lowers the amount of expertise needed to run a campaign and raises the number of targets one operator can handle. This dual-use character is the reason security researchers pay close attention whenever such software appears in real intrusions.
The tool was developed in China. That origin is a fact about where the software came from and does not, on its own, say who used it in these incidents, a distinction that matters because the question of attribution remains open.
Open availability is part of the concern. Software built for legitimate testing is often published so that companies can evaluate their own defenses, and anyone who obtains a copy can point it at systems they do not own. The line between a defensive tool and an offensive one is drawn by the person using it, not by the code.
Security teams have long dealt with this problem in the form of earlier frameworks that began as testing aids and later appeared in criminal campaigns. The arrival of AI-driven versions adds speed and autonomy to that familiar pattern.
4. What is the South Korea connection?
The US cybersecurity company CrowdStrike had earlier tied ARTEX to attacks on South Korean banks in late September 2026. GMO's report places the same tool on servers used against Japanese companies, which suggests that the software is being applied in more than one country in the region within a short span.
The timing is close. Breach disclosures by Japanese firms cluster in September and October 2026, a period that overlaps with the South Korean activity CrowdStrike flagged. Two separate security firms observing the same tool in neighboring countries during the same weeks gives the finding added weight, even though each report rests on its own evidence.
For defenders, the overlap is a practical signal. Indicators, behaviors and tactics documented in one country's incidents can help organizations elsewhere in the region look for similar traces in their own environments.
5. Who is behind the attacks?
That remains unanswered. GMO shared its findings with Japanese law enforcement, but the information it provided did not point to a specific actor. The account of the research does not name a group, a country or an individual responsible for the intrusions.
The distinction is important because the pattern of targets and timing might suggest coordination, and it is tempting to read it as the work of one organized group. The researchers' evidence, however, concerns the tool and the servers, not the people operating them. A widely available program can be run by many different parties, so the presence of ARTEX does not identify its user.
Until investigators connect the infrastructure to specific operators, the responsible reading is that the activity shows a method, namely AI-assisted attack tooling aimed at Japanese organizations, without establishing a culprit.
Attribution is typically the slowest part of any investigation. It requires technical evidence, cooperation between agencies and often information that only becomes available long after the first incidents. For now, the handover of GMO's findings to law enforcement is the first step in that longer process.
6. Why does it matter for Japanese companies and the financial sector?
The fact that nine of the roughly 20 targets are in one sector suggests that attackers may be going after weaknesses shared by firms in that industry. Common software, common suppliers or common practices can leave many companies exposed to the same technique, and an automated tool can test for such a flaw across a whole group of targets quickly.
The six additional infrastructures carry a broader message for retailers, schools, financial companies and media organizations: AI-assisted tooling is already pointed at their sectors. None of them can assume that the problem sits with some other industry.
Finance deserves particular attention. Japanese financial institutions appear among the targets on GMO's list, and ARTEX has previously been linked to attacks on banks in South Korea. For banks, payment companies and the crypto businesses that operate alongside them, the combination of a capable automated tool and a history of attacks on lenders is a reason to review their defenses.
Automation also changes the economics of defense. When the cost and skill needed to attack fall, the number of attempts rises, and security teams that rely on spotting occasional, hand-crafted intrusions may find themselves facing a steadier stream of automated probing. Faster patching, tighter monitoring and regular testing of one's own systems become more important in that setting.
There is a wider lesson for any business that holds sensitive customer data, including the digital asset firms that depend on strong security to keep user funds and personal information safe. Attacks that begin with an automated scan of exposed systems do not distinguish between a retailer and a trading platform; both are simply targets whose weaknesses can be found by the same software.
7. What can organizations do?
The practical response is largely familiar security hygiene applied with more urgency. Organizations can inventory the systems exposed to the internet, close or patch the weaknesses that an automated scan would find first, and make sure that unusual access to their data triggers an alert quickly enough to matter.
They can also use the same class of tools defensively. If an AI-driven tester can find a flaw in minutes, a company that runs it on its own network first has the opportunity to repair the problem before an attacker arrives. The technology that worries defenders in the hands of intruders can be a strong asset when it is used for authorized testing.
Sharing information helps as well. GMO's decision to publish its findings and to inform law enforcement gives other companies the chance to compare the indicators with their own logs, and the earlier warning from CrowdStrike about the South Korean attacks serves the same purpose.
Preparation also includes the human side of response. Teams that have rehearsed how to isolate affected systems, notify the right authorities and communicate with customers tend to recover faster than those improvising under pressure, and the cluster of Japanese disclosures shows how quickly such incidents can follow one another.
8. What happens next?
The next developments will probably come from law enforcement and from further research. Investigators who received GMO's findings may look for links between the infrastructure and specific operators, and other security firms may report additional victims or additional uses of ARTEX.
Observers will be watching for several things: whether more Japanese companies disclose breaches that fit the pattern, whether the six other AI-based infrastructures are tied to the same tool or to different ones, and whether any actor is formally identified. Until then, the picture is one of a capable AI testing tool turning up in real attacks across the region, with its operators still unknown.
For companies, the immediate task does not depend on that outcome. Whoever is responsible, the tool and the methods it automates are now documented, and organizations can act on that knowledge today.