Hackers in Japan Are Using Teams of AI Helpers to Break Into Networks

Security researchers have described three cyber attacks on organizations in Japan where artificial intelligence did a lot of the work. In one case, a single tool split the job across 44 AI agents, small programs that each handle a task, from scouting targets to stealing data. The findings come from TrendAI, the enterprise security division of Trend Micro.

What the researchers found

TrendAI studied three exposed server environments left behind by attackers. They held target lists, scripts, results, logs and command histories. Some were used mostly by hand by skilled operators. Others were AI platforms that automated the whole workflow.

The first attack used CyberStrikeAI, an open-source tool built for penetration testing, which is authorized hacking meant to find weak spots. It was previously used in a campaign against Fortinet FortiGate devices. This time it ran a repeatable process: scouting, testing for weaknesses, breaking in, staying hidden, stealing data and writing reports. It could also create forms and fake login pages in Japanese.

In a short period in October 2026, at least 19 organizations were tested, including at least 16 in Japan. Each had a report in the same format, which suggests a structured process rather than an operator's loose notes.

The other two cases

The second case was led by a human who used a large language model, the kind of AI behind chatbots, to speed up planning and tool building. One intrusion hit a mapping server left with default credentials. The attackers took land and business records, Windows credentials and email data over a 24-day campaign in August and September 2026.

In the third case, AI only helped write the operator's tools and scripts. This attacker went after devices from Cisco, Citrix, F5 and Yamaha, plus WebLogic servers, and ran at least six campaigns from January to October 2026.

Who is at risk

The targets were not limited to Japan. The first operator also went after organizations in Taiwan and Hong Kong. The third attacker's lists included organizations in the U.K., France, Italy, Qatar and Colombia.

TrendAI put it simply: AI has increased the attackers' workforce, but the work still consists of looking for unlocked doors. What changed is the ability to try thousands at once, adapt to a country's language and keep going without fatigue.

What is not known yet

There is no evidence that the three incidents share an operator. It is also unknown whether other regions were hit and which AI models were used.

What to do

• Change default passwords on any device or server that faces the internet.
• Install updates quickly on VPNs and other edge devices.
• Check which systems are exposed to the internet and remove any that do not need to be.