Chinese spyware tool LightSpy found in 13 countries

A Chinese-built spyware platform has evolved into a broader surveillance tool now operating in more than 13 countries, while a separate investigation has documented what researchers describe as the first fully autonomous AI-powered cyberattack campaign linked to a Chinese-speaking threat actor.

LightSpy Expands Global Reach

Arctic Wolf Networks said Wednesday that it detected a malicious software platform called LightSpy that enables paying customers to steal victims' personal information, including precise location data, audio recordings, chat records, camera footage, and screen recordings. The tool can also completely wipe a target's personal device, according to the firm.bloomberg

LightSpy was first reported in 2020 in connection with a watering-hole attack against Apple device users and has historically focused on the Asia-Pacific region. Arctic Wolf Labs identified an evolution in the campaign in 2024 when it documented the deployment of an advanced "DeepData" framework targeting Southern Asia. The latest findings, reported by Bloomberg, indicate the platform has expanded well beyond its original geographic focus.arcticwolf+2

AI-Powered Autonomous Attacks

Separately, Palo Alto Networks' ↗ 5.00% Unit 42 research team published findings in late July detailing how a Chinese-speaking threat actor used DeepSeek through the open-source Hermes Agent framework to conduct autonomous cyberattacks. The operation, triggered by a single command sent over Telegram, targeted more than 460 systems, scanning for vulnerabilities and attempting exploitation with minimal human intervention.reddit+3

Of the roughly 460 attempts, Unit 42 confirmed only three successful breaches — all involving data extraction from Citrix NetScaler instances through vulnerability CVE-2026-3055. Attempts against Langflow, n8n, and other platforms largely failed.reddit

The threat actor, operating under the pseudonyms "knaithe" and "KnYuan" and believed to be based in Zhuhai, China, also experimented with Claude Code and OpenAI's models, but safety guardrails from those providers blocked the malicious requests. DeepSeek, accessible via an open-source framework without client-side restrictions, proceeded without hindrance.bleepingcomputer+1

Implications for Defense

The incidents underscore the growing sophistication of China-linked cyber operations. While autonomous AI agents still face limitations in complex multi-step attacks, the Unit 42 case represents what BleepingComputer called "one of the initial concrete cases demonstrating that vendor-side safety measures can offer tangible defensive benefits". Organizations are advised to prioritize secure configurations, timely patching, and continuous monitoring to reduce exposure to both traditional spyware and AI-assisted threats.cybersecurity-insiders+2