The Ministry of Industry and Information Technology's cybersecurity threat and vulnerability sharing platform has identified a serious security backdoor risk in versions 2.1.91 to 2.1.196 of the AI programming tool Claude Code, developed by US-based Anthropic. The tool's built-in monitoring mechanism can transmit sensitive information such as geographic location and user identifiers to remote servers without user consent. It is recommended that relevant organizations and users immediately investigate, uninstall the affected versions, or upgrade to the latest secure version that has removed the backdoor code. Additionally, they should strengthen controls over outbound connection permissions and traffic monitoring for development tools within core business network segments.

Focus on the Trend: AI Supply Chain Security in the Wake of Alibaba's Ban

Major Companies Enforce Blanket Bans to Prevent Data Leaks

This incident is not an isolated security event but reflects potential risks in the supply chain security of large model development tools. According to reports, following a comprehensive internal assessment, Alibaba has added Claude Code, due to its implanted backdoor risk, to its high-risk software list. Since July 10, the company has completely prohibited its employees from using the tool in the office environment, recommending Qoder as an alternative instead. This move indicates that leading tech firms are beginning to adopt strict prohibition strategies against AI programming assistants with critical security vulnerabilities to prevent sensitive data leaks.

The Double-Edged Sword of AI: From Assistive Tool to Attack Vector

While AI tools enhance efficiency, their cybersecurity boundaries are increasingly blurred. A recent Google report noted the first observed instance of cyber attackers leveraging AI technology to develop "zero-day vulnerability" attack tools. Although Anthropic has previously "weakened" its main models to bolster defenses, third-party tools or specific versions can still become entry points for attacks. This highlights the new attack vectors potentially introduced by AI code generation and modification.

Regulatory Acceleration: Shifting from Passive Defense to Active Control

Regulatory bodies are accelerating efforts to improve AI supply chain security mechanisms. The National Financial Regulatory Administration has explicitly required financial institutions to establish supply chain security compliance management mechanisms for AI models and technical tools to prevent supply chain poisoning. The Ministry's current alert aims to push enterprises from a "passive defense" stance towards "active control," by restricting development tool permissions and enhancing traffic monitoring to build a safer AI application ecosystem.

Analysis: The AI Agent Paradigm Shift and Security Wake-Up Call

Claude Code, developed by Anthropic, is driving a shift in software development from passive tools towards autonomous Agent paradigms, thanks to its superior performance in long-text comprehension, complex reasoning, and coding that surpasses traditional Copilots. However, the monitoring mechanism built into versions 2.1.91 to 2.1.196 reveals a potential neglect of privacy protection in the pursuit of efficiency by AI models. This incident, similar in nature to the data leak risks posed by OpenClaw due to misconfigurations, underscores the urgency of balancing AI development with security.

The Government Work Report has included "AI+" initiatives for three consecutive years, and for the first time proposed creating new forms of the intelligent economy, while also emphasizing the construction of an AI governance framework. Therefore, the industry should prioritize enhancing data privacy and security protection capabilities. Users are advised to immediately investigate and upgrade versions while strengthening outbound connection controls in core network segments.

Investment Thesis: Equal Emphasis on Development and Security, Benefiting AI Governance and Cybersecurity

The security incident with Claude Code significantly reinforces the industry logic of "giving equal emphasis to development and security," providing strong positive momentum for the AI governance and cybersecurity sectors.

Policy: The Government Work Report's focus on "AI+" for three years and its new proposal for an intelligent economy, alongside explicitly including AI governance, highlights the importance of data privacy and security.

Industry: With the explosive popularity of AI agents like OpenClaw and the rapid adoption of Claude Code, demand for computing power remains robust. Concurrently, decisive actions by leading firms like Alibaba to ban risky tools signal an upgrade in supply chain security controls, accelerating the process of domestic substitution.

Regulation: Measures introduced by the National Financial Regulatory Administration and the Ministry of Industry and Information Technology are pushing financial institutions, government, and enterprises to establish strict AI supply chain security compliance mechanisms.

This series of actions will significantly increase market demand for AI security, data privacy protection, and domestically controllable AI tools, benefiting cybersecurity, domestic computing chips, and the development of an independent software ecosystem. It is recommended to focus on high-quality targets with core security technologies and autonomous control capabilities, seizing the structural investment opportunities arising from the standardized development of the AI industry.