Espionage by AI agents: China-linked hackers targeted more than 200 organisations
According to CrowdStrike, the group VAPOR PANDA left much of its attack work to AI agents based on DeepSeek – over eleven days and in at least 14 countries. Nine organisations were confirmed compromised.
Symbolic image · Photo: Coolcaesar / Wikimedia Commons, CC BY 4.0 (cropped and resized)
A hacking group attributed to China has, according to CrowdStrike's observations, used AI agents for a worldwide espionage campaign for the first time, the US security firm CrowdStrike reported on Thursday. The group, which CrowdStrike calls VAPOR PANDA, used the agent framework ARTEX: a main agent directs planner and worker agents that proceed from initial access through host reconnaissance to lateral movement in the network. The language model was mainly China's DeepSeek, supplemented by a second commercial model that CrowdStrike could not identify with certainty.
Starting on 17 October, the group targeted more than 200 organisations in at least 14 countries within eleven days, including foreign ministries in Latin America, Europe, Asia and the Middle East, as well as US companies in the semiconductor, defence, pharmaceutical, nuclear and financial sectors. At nine organisations, CrowdStrike confirmed that the attackers had gained access and carried on – among them telecoms providers in the US, Belgium and the Philippines and an interior ministry in the Middle East. In a single intrusion, the planning agents consumed around 1.4 billion input tokens.
Among other things, the attackers exploited poorly secured services and access keys from older data leaks that had apparently never been changed. CrowdStrike distinguishes the campaign from the ARTEX attacks Microsoft reported a little over a week ago under the name Storm-0940.