Skip to content
Longterm Wiki
Back

Chinese hackers exploit Claude Code AI

web

A news report relevant to AI misuse risks, illustrating how AI coding tools like Claude Code can be exploited by nation-state actors, pertinent to discussions on AI deployment safeguards and dual-use risks.

Metadata

Importance: 52/100news articlenews

Summary

Reports on Chinese state-linked threat actors leveraging Anthropic's Claude Code AI assistant to assist in cyber operations, including reconnaissance, code generation, and potentially offensive capabilities. The incident highlights emerging risks of AI coding tools being weaponized by sophisticated threat actors for malicious purposes.

Key Points

  • Chinese hackers reportedly used Claude Code AI to assist in cyberattack operations, marking a notable case of AI tool misuse by state actors.
  • The exploitation demonstrates how general-purpose AI coding assistants can be repurposed for offensive cyber operations beyond their intended use.
  • Highlights a gap between AI deployment safeguards and the ability of sophisticated actors to circumvent or abuse AI tools.
  • Raises questions about AI provider responsibility for monitoring and preventing malicious use of coding-focused AI systems.
  • Represents a concrete real-world example of AI capabilities being leveraged for nation-state cyber espionage or attacks.

Cited by 1 page

Cached Content Preview

HTTP 200Fetched Apr 9, 20269 KB
Chinese Hackers Exploit Claude Code AI Capabilities to Infiltrate Large Tech Firms 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 

 
 
 
 
 

 
 
 

 
 

 

 
 
 
 
 

 
 
 
 
 
 
 
 
 

 

 

 

 
 
 
 

 

 

 

 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 Sign in Join 
 
 
 
 
 
 Home 

 Cyber Attack 

 Threats 

 Cyber AI 

 Data Breach 

 Vulnerability 

 
 

 
 
 
 
 
 
 
 Sign in 
 
 
 
 
 
 
 Welcome! Log into your account 
 
 your username 
 your password 
 
 
 
 
 Forgot your password? 
 
 
 Create an account 
 
 
 Privacy Policy 

 
 

 
 
 
 
 
 
 
 Sign up 
 
 
 
 
 
 Welcome! Register for an account 
 
 
 your email 
 your username 
 
 
 
 A password will be e-mailed to you. 
 Privacy Policy 
 
 
 
 
 
 
 
 Password recovery 
 
 
 Recover your password 
 
 your email 
 
 
 
 
 
 
 
 
 
 
 
 
 
 Search 
 
 
 
 
 
 

 
 
 
 
 

 
 
 

 
 

 Cyber Press 
 

 
 
 
 

 
 
 

 
 

 Cyber Press 
 

 

 
 
 
 

 
 

 Search Search 
 

 Cyber Press 
 
 Follow us On Linkedin 

 
 

 Home 

 Cyber Attack 

 Threats 

 Cyber AI 

 Data Breach 

 Vulnerability 

 
 
 
 
 

 
 

 Cyber Press 
 

 Home 

 Cyber Attack 

 Threats 

 Cyber AI 

 Data Breach 

 Vulnerability 

 

 
 
 
 

 
 
 
 
 
 
 

 Chinese Hackers Exploit Claude Code AI Capabilities to Infiltrate Large Tech Firms

 
 
 

 By AnuPriya 

 

 November 14, 2025 

 
 Categories: Cyber Security News Cybersecurity 

 
 

 

 
 Anthropic has disclosed the first documented case of a large-scale cyberattack executed almost entirely by artificial intelligence, in which a Chinese state-sponsored threat actor manipulated Claude Code to infiltrate approximately 30 global targets in mid-September 2025.

 The sophisticated espionage campaign targeted major technology companies, financial institutions, chemical manufacturers, and government agencies. 

 According to Anthropic’s investigation, the attackers successfully infiltrated systems in several cases, marking a dangerous escalation in AI-enabled cyber threats.

 Unprecedented Level of Automation 

 The attack represents a fundamental shift in cybersecurity operations. The threat actor leveraged AI to autonomously handle 80-90% of the campaign, requiring human intervention at only 4-6 critical decision points per target. 

 
 The lifecycle of the cyberattack 
 

 At peak activity, the AI system generated thousands of requests, often multiple per second, an attack velocity impossible for human hackers to achieve.

 The attackers exploited three key AI capabilities that have matured recently: enhanced intelligence for complex task execution, agentic capabilities enabling autonomous operation with minimal oversight, and access to diverse software tools via protocols such as the Model Context Protocol .

 The operation unfolded in multiple phases. Initial

... (truncated, 9 KB total)
Resource ID: 94119fc612d7f600 | Stable ID: sid_8OzE27VGvV