China Flags Security Backdoor in Anthropic’s Claude Code Tool

Chinese Regulator Raises Alarm Over AI Coding Assistant

China’s National Vulnerability Database (NVDB) issued a stark warning Wednesday about an alleged security backdoor embedded in versions of Anthropic’s Claude Code, an AI-powered coding assistant. The cybersecurity platform, which operates under China’s Ministry of Industry and Information Technology, claims the tool could transmit sensitive information including users’ locations and identity-related identifiers back to Anthropic’s servers without consent. The allegations mark the latest flashpoint in escalating tensions between Chinese regulators and American AI companies over data security and access restrictions.

Claude Code functions as an AI coding agent that generates computer code, debugs software, and reviews code based on user prompts. The San Francisco startup Anthropic officially blocks users and companies in China and other nations it deems adversarial from accessing its products. However, users in the country continue to access these tools through VPN services or third-party proxy solutions, creating a persistent enforcement challenge for the company.

The NVDB stated on its website that it had recently “detected that the AI coding tool Claude Code contains security backdoor risks, posing a severe threat.” The regulator advised relevant institutions and users “to conduct a comprehensive check immediately” and “promptly uninstall or upgrade to the latest secure version” from which the controversial backdoor code has been removed. Additionally, it urged organizations to strengthen network traffic monitoring to prevent unauthorized leakage of sensitive data.

Alibaba Bans Tool Amid Security Concerns

The warning carries significant weight in China’s tech sector, where regulatory compliance is mandatory. Chinese tech giant Alibaba told employees last week that the use of Claude Code would be banned starting July 10 due to security concerns, according to people familiar with the matter. The move demonstrates how quickly major Chinese corporations respond to government cybersecurity warnings, particularly regarding foreign technology platforms.

The relationship between Anthropic and Alibaba has been fraught with tension beyond this latest controversy. Anthropic has previously accused Alibaba of reverse-engineering its AI models to mimic their abilities in a process known as “distillation.” This practice involves using a sophisticated AI model to train a separate, often simpler model to replicate its performance without directly accessing the original model’s architecture or training data.

The allegations about Claude Code first emerged in specialist tech media last week, prompting responses from Anthropic personnel. Claude Code engineer Thariq Shihipar addressed the reports in an X post, providing context for the controversial mechanism. “This is an experiment we launched in March that was meant to prevent account abuse from unauthorized resellers and protect against distillation,” Shihipar wrote. He added that the team had since developed stronger mitigations and had been planning to remove this feature.

Company Frames Mechanism as Anti-Abuse Measure

Anthropic told AFP Wednesday that the mechanism checked a device’s timezone and whether a request was routed through a domain tied to an unsupported region or a known problematic entity. The company characterized this as a standard way to detect fraud and abuse, positioning the controversial code as a protective measure rather than a security vulnerability. However, Anthropic has not issued a formal public statement addressing the Chinese regulator’s specific allegations.

Shihipar indicated that the company planned swift action to address concerns. “The team has landed stronger mitigations since then and we’ve actually been meaning to take this down for a while… this should be fully rolled back in tomorrow’s release,” he explained. The promise of a rapid rollback suggests Anthropic recognized the mechanism had become more problematic than beneficial, regardless of its original intent.

The incident illuminates the complex challenges AI companies face in global markets. They must navigate conflicting demands that include preventing abuse and protecting intellectual property. At the same time, they must maintain user trust and comply with diverse regulatory environments across different jurisdictions. These competing priorities often create technical and ethical dilemmas with no perfect solutions.

Geopolitical Tensions Shape AI Regulation

The controversy surrounding Claude Code highlights tensions at the intersection of cybersecurity, geopolitical conflict, and AI development. As artificial intelligence tools become more integral to software development and business operations, several critical questions arise. These include concerns about data collection, user privacy, and national security, which grow increasingly urgent as AI capabilities expand.

Chinese regulators have grown increasingly vigilant about foreign technology products operating within their borders, particularly those from American companies. The NVDB’s warning reflects broader concerns about data sovereignty and the potential for foreign entities to collect intelligence on Chinese users and organizations. These concerns mirror similar anxieties expressed by Western governments about Chinese technology companies operating in their markets.

The perception of mechanisms like those in Claude Code varies significantly across regulatory frameworks. What Anthropic describes as a legitimate anti-abuse measure, Chinese regulators characterize as an unacceptable security risk. This divergence depends heavily on perspective, institutional trust, and the specific regulatory framework governing technology in each jurisdiction. Neither view is objectively incorrect; rather, they reflect fundamentally different priorities and threat models.

Broader Implications for AI Industry

The Claude Code controversy arrives amid broader global debates about AI governance, cross-border data flows, and technology sovereignty. Companies developing advanced AI systems face mounting pressure to balance innovation with security, accessibility with abuse prevention, and global reach with local compliance. These pressures intensify as AI capabilities advance and deployment scales expand across industries and geographies.

For users in China who rely on VPNs to access restricted foreign AI tools, the NVDB warning creates uncertainty about continued use. Organizations must weigh the productivity benefits of advanced AI coding assistants against potential security risks and regulatory consequences. The July 10 deadline for Alibaba employees represents just one example of how companies are responding to these competing pressures.

The incident also underscores the enforcement challenges facing AI companies that attempt to restrict access by geography. Technical measures to detect and block users from specific regions often prove imperfect, creating gray areas where determined users can still access services despite official prohibitions. These workarounds complicate both security measures and regulatory compliance for all parties involved.

As Anthropic moves to remove the controversial mechanism from Claude Code, the episode will likely inform how the company and its competitors approach similar challenges in the future. The balance between protecting intellectual property from distillation, preventing account abuse, and maintaining user trust remains one of the most difficult problems facing AI companies operating across adversarial jurisdictions in an increasingly fragmented digital landscape.