Alibaba is reportedly preparing to ban the use of Anthropic’s Claude Code across its internal environments starting July 10. This decision comes in light of allegations that the AI-powered coding assistant has a covert detection mechanism resembling a backdoor.
The news, first reported by the Chinese financial outlet Yicai and later confirmed by Reuters, has not been officially acknowledged by Alibaba. However, it highlights the growing tensions among major AI vendors amid concerns over model security, data exfiltration, and competitive intelligence gathering.
Claude Code, Anthropic’s command-line-based coding assistant, has rapidly gained popularity among enterprise developers for its ability to generate, debug, and optimize code directly within terminal environments.
Its growing presence in enterprises makes the alleged security concerns particularly significant, especially for large organizations like Alibaba that manage extensive cloud and AI infrastructure.
This reported ban arises at a time when both Anthropic and Alibaba have recently exchanged accusations regarding model distillation practices and unauthorized data extraction.
The controversy began with a June 30 Reddit post by a user named “LegitMichel777,” who claimed to have reverse-engineered Claude Code while trying to restore a disabled remote control feature.
According to the technical analysis shared by the user and later summarized by multiple security-focused outlets, versions of Claude Code since 2.1.91 (released on April 2) allegedly contain hidden logic that inspects system-level configurations.
Specifically, the tool reportedly checks whether a user’s proxy settings or system timezone aligns with entries in two concealed lists associated with Chinese corporate networks and AI research entities, including Alibaba, Baidu, ByteDance, and Moonshot AI.
Instead of directly transmitting telemetry, the mechanism allegedly encodes detection results by subtly modifying internal system prompts, altering date formats and substituting punctuation characters.
If validated, this could represent a new form of covert environmental fingerprinting designed to evade traditional detection methods while enabling behavioral tracking or policy enforcement.
Anthropic has not released a formal public statement addressing these allegations. However, a member of the Claude Code development team reportedly acknowledged the feature on social media, describing it as an anti-abuse mechanism intended to detect account reselling and large-scale attempts at model distillation.
The developer indicated that this functionality would be removed in an upcoming release, with reports suggesting remediation efforts were already underway as of July 1. Based on this timeline, the feature may have been active for approximately three months.
This dispute occurs against a broader backdrop of increasing restrictions across AI platforms aimed at reducing model distillation and unauthorized access.
In a June 10 letter to U.S. lawmakers, Anthropic alleged that entities linked to Alibaba’s Qwen AI division orchestrated a large-scale campaign involving nearly 25,000 fraudulent accounts, generating over 28 million interactions with Claude models within six weeks. Alibaba has not publicly commented on these claims.
Despite extensive discussions, no independent security audit has yet confirmed the existence or intent of the alleged backdoor. The absence of official statements from both companies leaves important questions unresolved about whether the mechanism poses a security risk, serves as a defensive anti-fraud control, or is a misunderstood feature.
If implemented, Alibaba’s ban would be one of the first enterprise-level restrictions imposed specifically over suspected covert functionality in an AI coding tool, potentially setting a precedent for how organizations evaluate trust in AI-assisted development platforms.
Interact with Cyber Threats in Windows, Linux, macOS VMs to Trigger Full Attack Chain - Analyse Malware & Phishing with ANY RUN
