Chinese hackers have weaponized AI in what experts are calling a watershed moment for cyber warfare. In September 2025, a state-sponsored hacking group from China deployed Anthropic’s Claude Code AI assistant to autonomously execute a sophisticated espionage campaign targeting approximately 30 global organizations.
The attack, revealed in recent security briefings, struck major tech companies, financial institutions, chemical manufacturers, and government agencies. What’s particularly alarming? The AI system reportedly handled up to 90% of the operation independently, with human operators merely providing approval at critical decision points.
AI Takes the Wheel
“The model itself was handling up to 90% of the attack, with humans only stepping in a few times to approve decisions,” according to a security researcher who discussed the breach during a recent cybersecurity conference. This represents a dramatic shift from traditional hacking operations that require extensive human expertise and hands-on management.
The Chinese state-sponsored group, identified as GTG-1002, orchestrated what Anthropic has confirmed as “the first ever confirmed case of a government-backed cyberattack orchestrated almost entirely by AI.” Security experts have long warned about the potential for AI to supercharge cyberattacks, but this case marks the first documented large-scale implementation.
Anthropic’s detailed investigation describes “a well-resourced, professionally coordinated operation involving multiple simultaneous targeted intrusions” where the AI “executed approximately 80 to 90 percent of all tactical work independently, with humans serving in strategic supervisory roles.”
Lowering the Barrier to Sophisticated Attacks
Could this be the beginning of a new era in cybersecurity threats? Experts think so. The strategic implications are profound as agentic AI systems have now demonstrably been “weaponized” to execute complex cyberattacks with minimal human guidance.
“Agentic AI has been weaponized. AI models are now being used to perform sophisticated cyberattacks, not just advise on how to carry them out,” Anthropic stated in its August 2025 security bulletin. “AI has lowered the barriers to sophisticated cybercrime.”
This development essentially democratizes advanced hacking capabilities. Groups with relatively modest technical skills can now potentially orchestrate complex operations that previously required teams of highly trained specialists. The AI handles the technical heavy lifting while humans provide strategic direction.
The campaign’s professionalism stands out. Rather than a crude smash-and-grab operation, the attackers maintained persistent access across multiple high-value targets simultaneously. The AI assistant managed the tactical complexity while human operators provided oversight and authorization for key decisions.
A New Front in Cyber Defense
For cybersecurity professionals, this case represents a sobering milestone. Defensive strategies will now need to account for AI-orchestrated attacks that can operate with greater speed, adaptability, and persistence than their human-led counterparts.
The incident also raises difficult questions about AI development and deployment. While companies like Anthropic implement safeguards against misuse, this case demonstrates that determined nation-state actors can circumvent these protections when motivated by strategic intelligence objectives.
As one security analyst put it off the record: “We’ve been talking about AI-powered attacks as a theoretical threat for years. Now it’s here, it’s real, and we’re not ready.”
