Hackers Use Claude, Qwen and DeepSeek AI Agents in Attacks on Asian Organisations
Researchers Detect Growing Use of Artificial Intelligence as Operational Tools in Cyberattacks Targeting Governments and Other Institutions
Chinese-speaking threat actors have reportedly begun using AI agents powered by Claude, Qwen and DeepSeek as part of cyber intrusion campaigns targeting organisations across Asia. Researchers say the activity represents a growing shift in which artificial intelligence is being incorporated directly into offensive cyber operations rather than being used only for basic tasks such as drafting or translation.
The latest campaign is reportedly separate from an earlier operation identified in July that involved Claude Code and DeepSeek. The newer activity has expanded the range of AI models being used and reportedly targeted government agencies, political organisations, educational institutions and industrial organisations in several Asian countries.
Researchers identified several organisations reportedly affected by the campaign, including Taiwan’s Kuomintang Party History Archives, Indonesia’s Ministry of Foreign Affairs, government and education organisations in mainland China, and industrial systems in Vietnam. One reported breach involved a government office environment where attackers obtained access to systems and collected sensitive records.
At the centre of the operation was reportedly an AI-agent framework designed to divide larger objectives into specialised tasks. Investigators said the framework could work with different AI models, allowing operators to switch between Claude, Qwen and DeepSeek without changing the overall task structure. This development has raised concerns about how increasingly capable AI agents could affect the scale and speed of future cyber threats.
Security researchers say the campaign demonstrates why organisations need stronger cybersecurity practices as AI becomes more accessible to both defenders and attackers. Recommended defensive priorities include keeping internet-facing systems patched, monitoring unusual activity and strengthening detection capabilities. Researchers have attributed the infrastructure to a Chinese-speaking operator with moderate confidence but said there is not enough evidence to identify a specific state-sponsored group.
