Chinese hackers use AI tool to go after 200+ organizations at once

A Chinese state-linked hacking group has used an AI-powered tool to target more than 200 organizations around the world at the same time. Security firm CrowdStrike says the group, which it calls Vapor Panda, began the campaign on October 17, 2026. The tool is called ARTEX.

What happened

The targets included foreign ministries in Latin America, Europe, Asia and the Middle East. They also included semiconductor, defense and think tank organizations in the U.S., plus telecom providers in several regions.

CrowdStrike confirmed that the attackers got in and carried out follow-up work at nine organizations. They installed a malicious program called BLOODALCHEMY, then guessed passwords, scanned networks and mapped what was inside. The group also tried password guessing against at least 12 more organizations, including U.S. think tanks and government bodies in Europe.

How the tool works

ARTEX is a penetration testing framework, a kind of software normally used to probe systems for weaknesses. Here it was used for real attacks. It works with AI models, including DeepSeek and a paid commercial model that CrowdStrike has not identified.

• Planner and worker AI agents coordinate to move from first access to scouting and spreading inside a network.
• A human operator oversees the work, directing agents to do things like steal administrator passwords.
• The agents tried many attack paths in parallel and switched approaches when one failed.

CrowdStrike says this lets a small team pursue far more targets than manual hacking would allow. In one break-in alone, the planning agents used about 1.4 billion input tokens. Tokens are the small chunks of text an AI reads, so the number hints at how much automated activity took place.

Other groups in the report

CrowdStrike also described four newly documented China-linked groups: Hollow Panda, Overcast Panda, Titan Panda and Veiled Panda. Overcast Panda has installed spyware on devices of people traveling in China, using direct physical access. Titan Panda attacked law firms through a flaw in legal billing software called Intapp Time Server, with attacks first detected in April 2026.

What it means for you

The ARTEX attacks relied on well-known methods, such as password guessing, alongside the AI. The AI did not invent a new kind of attack. It made old ones faster and cheaper to repeat across many victims. Any business could end up on a long target list, not only governments and defense firms.

Basic habits still block a lot of this. Strong, unique passwords make guessing harder. Prompt software updates close holes that attackers look for, like the one Titan Panda used. CrowdStrike has not said how the group chose its targets or how the break-ins began at each one.