Starting July 10, 2026, Alibaba is imposing a categorical ban on the use of all Anthropic products within its corporate environment. The restrictions cover Claude models — Sonnet, Opus, and Fable — as well as the developer tool Claude Code. Employees are required to switch to the proprietary Qoder platform as a mandatory alternative. This decision is not merely an internal policy but a direct response to systemic security risks uncovered during an independent audit.

Hidden Mechanisms in Claude Code: What Analysts Discovered

A few days before the ban announcement, the developer community identified suspicious data collection features in Claude Code version 2.1.91 (released in April 2026). The tool scanned the user's local environment in the background: checking the time zone and proxy server addresses. Special attention was paid to keywords related to Chinese cloud providers, AI companies, and access resale services. When triggers matched, Claude Code altered the date format and punctuation in the system prompt — without sending explicit telemetry. This masking raised serious concerns within the professional community.

Tariq Shihab, a representative of the Claude Code team at Anthropic, called this feature "an experiment launched in March," allegedly aimed at combating unauthorized account resale and protecting against model distillation. He promised to remove the mechanism in the next release. However, for Alibaba, this "experiment" was the last straw, confirming the inadmissibility of using Anthropic products in corporate networks.

Geopolitical Context and Escalating Tensions

Alibaba's ban is a logical continuation of a long-standing conflict. As early as September 2025, Anthropic tightened its sales policy for unsupported regions, explicitly naming China and Hong Kong. The company justified this with risks of forced data disclosure to intelligence agencies and the threat of using Claude to develop Chinese AI systems. In June 2026, Anthropic sent a letter to the U.S. Senate, accusing operators linked to Alibaba of the largest known attack on Claude model distillation. According to my sources, this appeal was the direct trigger for the internal ban at Alibaba.

The Shadow Market and the Real Value of Claude

Despite strict restrictions, demand for Claude in China remains high. Accounts are actively sold on Taobao, Xianyu, and Telegram, with intermediary stations using servers in Anthropic-supported countries ensuring stable operation. Zilan Qian, a researcher at the Oxford China Policy Lab, rightly noted that Chinese models still lag behind American ones by six to nine months, especially in code-writing tasks. This explains why developers in China prefer Claude Code and OpenAI Codex over local alternatives, despite all the risks.

My analysis: Alibaba's ban is not an isolated incident but a symptom of a deep rift in the global AI industry. On one hand, we see the technological dependence of Chinese developers on Western models; on the other, growing paranoia around data security. Anthropic's experiment with background data collection, albeit "harmless," shattered the remaining trust. In the short term, this will impact the productivity of Chinese teams, but in the long term, it will accelerate the development of local alternatives, such as Alibaba's Qoder. The AI market is ultimately fragmenting, and those who can offer not just a powerful model but a secure and sovereign ecosystem will win.