Alibaba Blocks Claude Code as AI Trust Becomes a Developer Tool Problem
By Suad Seferi ·
Alibaba has reportedly told employees to stop using Anthropic’s Claude Code, turning a dispute between two AI giants into a wider question for developers: how much trust should companies place in AI coding tools that can access their systems? The move follows reports that Alibaba classified Claude Code as a high-risk tool and instructed staff to remove it from company devices. Employees were reportedly directed toward Qoder, Alibaba’s own AI coding assistant. The dispute began after researchers claimed Claude Code contained mechanisms that could identify users connected to China through signals such as time zone, proxy settings and network indicators. Anthropic has pushed back against claims that the feature was spyware, saying it was linked to anti-abuse efforts, including attempts to prevent unauthorized reselling and model distillation. For Alibaba, however, the risk appears to be less about one technical feature and more about control. AI coding agents are no longer simple autocomplete tools. They can read repositories, inspect files, run commands and influence production workflows. That makes transparency over what they collect, where they send data and how they identify users a core enterprise concern. The ban also comes after Anthropic accused China-linked actors, reportedly connected to Alibaba, of using thousands of fraudulent accounts to extract knowledge from Claude through large-scale model distillation. Alibaba has denied wrongdoing. Together, the two incidents show how fast the AI race is moving from model performance into security, sovereignty and trust. For companies in the Balkans, the lesson is practical. Before adopting AI coding assistants, organizations should ask where code is processed, what metadata is collected, whether logs are retained, and which tools are approved for internal use. Claude Code may remain a powerful tool for developers, but Alibaba’s decision shows that in enterprise AI, capability is no longer enough. The next competition is about who controls the infrastructure behind the intelligence.