Alibaba Bans Anthropic's Claude Code Over Security Risks
Alibaba has announced a ban on Anthropic‘s Claude Code tool for all employees, effective . The move follows internal security assessments flagging the coding assistant as high-risk software due to potential backdoors embedded in the tool. Chinese publication Yicai first reported the decision, which bars Alibaba staff from using Claude Code in office environments. The company is directing employees to Qoder as an alternative.

What the Backdoor Claim Says

According to Dutch publication Techzine, Alibaba has accused Anthropic of embedding a backdoor into Claude Code. The alleged mechanism was first highlighted by a Reddit user who claimed to have reverse-engineered the tool and published detailed findings on the platform.

Starting with version 2.1.91, released , Claude Code reportedly checked proxy configurations and time zones against hidden lists of Chinese corporate networks and AI labs. The monitored organizations included Alibaba, Baidu, ByteDance, and Moonshot AI. The Reddit post alleged that the code used XOR-obfuscation with the key 91 to hide this logic within the Claude Code binary, and that Anthropic omitted any mention of this check from the version 2.1.91 release notes.

Anthropic’s Response

Anthropic has not issued a formal statement addressing the broader claims. However, a Claude Code team member did respond to the initial Reddit post, stating that the mechanism was intended to combat account sales and model distillation. The team member promised the feature would be removed in the next release but did not directly address accusations about surveillance or data exfiltration.

Escalating US-China AI Tensions

This controversy builds on an already contentious relationship between the two companies. In June, Anthropic accused Alibaba’s Qwen lab of operating nearly 25,000 fraudulent accounts to scrape Claude outputs, generating 28.8 million interactions without authorization. Alibaba’s decision to ban Claude Code internally is widely viewed as a counter-move in this broader conflict.

The back-and-forth accusations reflect deepening distrust between US and Chinese AI firms as both nations compete for dominance in the space. Alibaba’s public ban signals both a security concern and a geopolitical statement.

What Remains Unclear

Several key questions remain unanswered:

  • Did Claude Code transmit any actual user data before version 2.1.91 or since?
  • What specific “distillation” or “account sale” risks was Anthropic trying to prevent?
  • Will Anthropic release a detailed timeline of the feature’s lifecycle and any data collection that occurred?

The lack of a formal, comprehensive statement from Anthropic leaves the tech industry without clarity on the scope and impact of the alleged backdoor. The company now faces pressure to address not just Alibaba’s concerns, but broader questions about transparency in AI tool development.

As of now, the tech industry awaits further developments, particularly any official response from Anthropic detailing the next Claude Code release and how it will address these security concerns.

Follow Hashlytics on Bluesky, LinkedIn, Telegram and X to Get Instant Updates

RELATED ARTICLESMORE FROM AUTHOR