Chinese state-sponsored hackers have deployed **ARTEX AI**, an open-source, multi-agent AI penetration-testing framework, in a widespread cyberespionage operation. The campaign, which began on October 17, targeted government ministries, telecommunications providers, and organizations in the pharmaceutical, semiconductor, defense, nuclear, and financial sectors across 14 countries. ### AI-Driven Attacks Scale Globally Researchers at **CrowdStrike** attribute the campaign to the threat actor **Vapor Panda**. They note that this operation signifies "an escalation in China-nexus use of AI agents, from single intrusions to multiple simultaneous engagements directed by a human operator." At least nine organizations have been compromised, with an additional 12 high-value targets subjected to password spraying and credential scanning activities. Confirmed victims include telecommunications providers in the United States and Belgium, government entities in Greece, the Middle East, and Latin America, and pharmaceutical firms in the U.S. ### How ARTEX AI Facilitated Intrusions In an attack against a Greek government entity, the **ARTEX AI** agents successfully forged an administrator token and accessed mailboxes by exploiting authentication keys derived from a 2022 mail-server directory export. **ARTEX AI**, originally developed by Chinese developers as an open-source project, later became proprietary after its use in real-world attacks, including a previous campaign against South Korean financial institutions. Beyond **ARTEX AI**, **Vapor Panda** also incorporated **DeepSeek** and another unidentified commercial language model to orchestrate planning and worker agents across multiple intrusions concurrently. **CrowdStrike** observed that the attackers issued instructions in Chinese and intervened at crucial decision points. The AI agents were capable of performing reconnaissance, exploiting vulnerabilities, deploying web shells, obtaining credentials, and dynamically pursuing alternative attack paths when initial attempts failed. Notably, the operators explicitly instructed the agents to bypass Windows machines protected by **CrowdStrike**'s **Falcon** endpoint protection solution, focusing instead on unprotected hosts. In one intrusion, researchers estimated the AI agents consumed over 1.4 billion input tokens. **CrowdStrike** also discovered the deployment of **BLOODALCHEMY**, a backdoor active since 2023, against government entities in Latin America and the Middle East during this campaign.  ### More ARTEX AI Cases Emerge The Dutch Institute for Vulnerability Disclosure (**DIVD**) has provided further details regarding a cyberattack it suffered in September 2026, which it attributed to an AI agent. The **DIVD** confirmed that **ARTEX AI** was the tool used in the autonomous breach of its network. According to **DIVD**'s **LinkedIn** post, the attack achieved initial access and escalated to root privileges within a mere 12 minutes. This rapid compromise involved chaining a **Zammad** zero-day vulnerability with a second zero-day for local privilege escalation. The AI agent executed numerous tasks simultaneously, including reading a compromised mailbox, scanning the internal network, extracting login tokens, and attempting password resets. **DIVD** documented over 450,000 requests and 2,500 scripts associated with the attack, describing the activity as "super loud and messy."