Chinese tech giant Alibaba has reportedly told employees to stop using Anthropic's AI coding tool "Claude Code" internally, citing alleged backdoor security risks, with the ban set to take effect in workplace environments from July 10, 2026.
July 3, 2026 · AI Governance
Alibaba Moves to Ban Anthropic's Claude Code Over Alleged "Backdoor"
An internal directive reportedly bars the AI coding tool in workplace environments from July 10, citing security risk from a user-identification feature — as the two firms clash over an alleged model-distillation attack.
Jul 10
Ban takes effect in workplace environments
Qoder
Alibaba's in-house tool recommended as replacement
Apr–Jun
Window of the disputed activity in 2026
The distillation dispute, drawn to scale
Anthropic's claim of its "largest-ever distillation attack" tied to Alibaba's Qwen lab — the scale of accounts vs. interactions.
~28.8M
interactions
Roughly 1,150 interactions per account, aimed at software-engineering & agentic reasoning capabilities.
Anthropic's defense
The feature is legitimate fingerprinting — inspecting time zone and proxy data to prevent unauthorized resellers and model extraction. No data transmission, they say.
China-side concern
Inserting hidden markers when networks match Chinese-linked labs is seen as a "backdoor" that violates privacy and discriminates against Chinese users.
The contradiction
Alibaba Cloud still officially sells a paid Claude Code integration — even as its internal security team moves to ban the same tool.
Neither company has confirmed the ban. Claude Code remains widely used — averaging ~20 hours per week per user across ~400,000 analyzed sessions — leaving the exact scope of both the detection feature and the restriction unconfirmed.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…