Hacker groups linked to China are scaling up their operations with the help of ARTEX, an AI-driven penetration testing framework — a category of software normally used by security teams to probe their own systems for exploitable weaknesses. According to CrowdStrike, the group tracked as Vapor Panda has used the tool since October 17, 2026, in attacks against more than 200 organizations worldwide. Much of the work is carried out by AI, while humans retain the final decision-making.

Planner and worker agents pursue many paths at once

ARTEX is built around a main agent that supervises “planner” agents, which design attacks, and “worker” agents, which execute them. The agents share findings as they scan systems, exploit vulnerabilities, and move through networks. CrowdStrike reported that Vapor Panda combined DeepSeek with an unidentified commercial AI system. In a single attack, the planner agents consumed 1.4 billion input tokens — the units of text that AI models process.

The campaign reached at least 14 countries, with foreign ministries and telecom providers among the targets. In the United States, attackers went after 19 organizations spanning the pharmaceutical, semiconductor, defense, nuclear, and financial sectors.

Scale did not guarantee success. CrowdStrike confirmed post-exploitation activity at only nine companies, including the deployment of the BLOODALCHEMY backdoor. Human operators then took over, creating web shells, extracting passwords, and locating unprotected machines.

A second group scales up, from ConnectWise to Microsoft 365

ARTEX has also been used by Storm-0940, another hacking group linked to China. According to Microsoft, the group began using the tool as early as September 2026, targeted hundreds of organizations, and gained privileged cloud access in under three hours in several cases. Microsoft also tied the group to the ConnectWise Cloud Backup compromise.

The volume of activity was substantial: Microsoft tracked more than 151,000 failed logins between October 10 and 14 against 327 accounts across eight companies, peaking at 2,900 attempts per minute. No successful login was observed.

The ConnectWise compromise was a separate strand of the group’s activity. Following that breach, attackers also attempted to change cloud-application credentials, as Cryptopolitan previously reported.

A financially motivated actor hits South Korean banks

A third ARTEX campaign pursued a different motive. In late September 2026, attackers used the tool against South Korean financial institutions, according to Zscaler, and stole personal information belonging to roughly 67,000 people. Zscaler assessed with moderate confidence that the attacker was Chinese-speaking and financially motivated, and it did not confirm any state sponsorship. CrowdStrike also found ARTEX files and Claude Code logs on attacker-controlled systems.

How hacking became agentic

The shift toward automated hacking — driven by agentic AI, systems that plan and carry out multi-step tasks with limited human supervision — began well before ARTEX. In 2025, Anthropic uncovered a Chinese state-sponsored campaign that used Claude Code against roughly 30 targets, with AI handling an estimated 80–90% of the work. Google has also documented how attackers are moving beyond simple AI prompts toward autonomous operations. Those earlier activities were separate from the ARTEX campaigns, but they illustrate how quickly AI-powered hacking is evolving.

What it could mean for banks, chipmakers, and cloud providers

The risks extend beyond the organizations already hit. Banks face potential customer-data breaches, chipmakers could lose valuable intellectual property, and cloud providers face a distinct concern: compromised application credentials that could expose multiple customers.

Industry figures show why faster attacks worry businesses. Verizon found that system intrusion accounted for 61% of breaches in its 2026 report, while Microsoft reported that exposed cloud workloads faced attacks within an average of 5.3 hours. The World Economic Forum found that 65% of large companies consider third-party and supply-chain vulnerabilities their biggest cybersecurity challenge, up from 54% in 2025.

As attacks grow faster and more automated, businesses may redirect security spending toward stronger identity controls, faster detection, and closer monitoring of suppliers.