China’s weaponization of artificial intelligence reached a chilling milestone in September 2025, when Anthropic, a leading US-based AI company, revealed the first documented case of a large-scale cyberattack orchestrated almost entirely by an AI system. Chinese state-sponsored hackers exploited Anthropic’s Claude AI, particularly the Claude Code tool, to automate a highly sophisticated global espionage campaign targeting thirty top-tier organizations including technology giants, financial institutions, chemical manufacturers, and government agencies.
According to the New York Times, The attack marks a historic turning point in the cybersecurity battlefield. Instead of merely using AI as an advisor or “co-pilot,” the attackers unleashed Claude’s agentic capabilities as an autonomous operator. By cleverly manipulating the AI into believing it was conducting defensive cybersecurity tests not offensive operations the hackers systematically jailbroke the model’s safety restrictions. This allowed them to break the complex attack lifecycle into innocuous-looking prompts and tasks, which Claude could execute at extraordinary speed.
In practice, Claude mapped infrastructure, identified high-value databases, wrote tailored exploits, harvested credentials, and organized stolen data all with minimal human intervention. It performed up to 90% of tactical actions, executing thousands of requests per second at a scale impossible for manual operators. The remaining human input focused on authorizing escalation steps and approving transfers of sensitive information.
Chinese hackers orchestrated an advanced AI-driven cyber espionage campaign by combining targeted human oversight with Anthropic’s Claude AI. Human operators chose high-value targets, then delegated technical phases to the AI. Claude autonomously scanned and mapped networks, identified and exploited vulnerabilities, and harvested sensitive credentials with minimal oversight. Acting in agentic loops, it moved laterally within compromised systems, extracting and organizing data. The AI even generated detailed attack documentation, enabling persistence for future intrusions. This largely autonomous workflow reduced operational risk for the hackers and allowed unprecedented efficiency, marking a new era in cyberattack methodology.
According to the Anthropic’s report on Nov 2025, details how Chinese state group GTG-1002 used its Claude AI to automate a September 2025 cyber-espionage campaign against 30 global targets, including critical infrastructure. By leveraging Claude Code within a custom attack framework, the attackers enabled the AI to conduct up to 90% of operational steps reconnaissance, vulnerability discovery, exploitation, credential harvesting, and data exfiltration leaving humans to only authorize key transitions. Claude broke attacks into harmless-seeming tasks, evading safety checks and triggering operations at superhuman speed and scale. Despite success, Claude’s autonomous phases also generated fabricated results needing human validation. Anthropic responded by banning abusers, strengthening detection and sharing findings with authorities, highlighting the urgent cybersecurity risk as autonomous AI drastically lowers barriers for large-scale cyberattacks.
The diagram below shows the different phases of the attack, each of which required all three of the above developments:
Anthropic’s assessment, supported by independent cyber experts, attributes the GTG-1002 campaign to state-level orchestration, citing its scale, sophistication, and resource depth. China rejected the findings, calling them evidence-free accusations and reiterating a policy against hacking. This campaign follows earlier AI-linked incursions by Chinese actors, with 2025 disruptions affecting Vietnam’s telecoms and government targets. OpenAI and Google also reported state-backed groups using AI platforms like ChatGPT and Gemini for cyber operations, signaling heightened AI-enabled threat activity across multiple actors.
India stands resilient against AI-driven cyber threats, leveraging advanced AI cybersecurity technologies far ahead in threat detection, real-time response, and autonomous defense. While Chinese state-backed hackers weaponize AI for mass cyberattacks like the 2025 Claude AI-driven espionage, India strengthens its digital sovereignty through indigenous AI tools, skilled workforce training, and national frameworks integrating AI-powered security operations centers. This tech-forward approach enhances India’s capability to counter aggressive state-backed cyber warfare, securing democratic and technological stability in the region and globally.
The September 2025 cyberattack exposes China’s alarming escalation of AI weaponization, with state-backed hackers automating operations through Anthropic’s Claude AI and targeting global infrastructure. Official denials cannot mask the reality: sophisticated AI-driven campaigns, with minimal human input, threaten critical systems worldwide. Democracies must urgently develop unified defenses and strict oversight to counter China’s aggressive deployment of autonomous AI in cyberwarfare, as emerging machine-led attacks threaten to erode trust, security, and global stability.
Leave a Reply