Alibaba Group has reportedly ordered its employees to stop using Anthropic's Claude Code for work. Blocking a tool or product at a company is not unusual on its own, but this is one of the first instances of a major global tech organization blocking a leading AI coding assistant outright. Alibaba points to security concerns, though the two firms were already on tense terms before this.
The Chinese conglomerate says Claude Code carries backdoor risks and, after a comprehensive evaluation, added it to a list of high-risk software with security vulnerabilities. From July 10 onward, staff will not be allowed to use Claude Code for official work, according to the South China Morning Post.
Alibaba has not confirmed the policy publicly, and it did not respond to a request for comment from the South China Morning Post. Chinese outlet Yicai first reported the ban, and Reuters later confirmed it independently through a separate source. Employees have reportedly been told to switch to Alibaba's own coding platform, Qoder.
One standout detail from Reuters' reporting is that enforcing geographic restrictions on individual developers is genuinely hard, since anyone can route traffic through US-based servers to mask their location. That likely explains why Alibaba chose a blanket internal ban rather than relying on Anthropic's own access controls to keep Claude Code out of Chinese hands.
What's notable is that Claude Code had gradually built a following among Chinese developers despite Anthropic's standing restrictions on access from China. Alibaba's notice effectively closes that door from the employer's side instead of waiting for enforcement from Anthropic. It reflects a company choosing to police its own staff rather than trust a foreign vendor's compliance systems.
Also read || 10 Technologies to Learn in 2026 for Career Growth
The hidden code behind the backlash
The dispute began when a Reddit user named LegitMichel777 posted on June 30, claiming to have reverse-engineered Claude Code while restoring a disabled remote-control feature. The post alleged that since version 2.1.91, released on April 2, the tool quietly checked whether a user's proxy settings or system timezone matched entries on two hidden lists.
Claude Code allegedly held a hidden list of names and networks tied to Chinese companies, including Alibaba, Baidu, ByteDance, and Moonshot AI. Public analysis of the code described region fingerprinting and watermarking of flagged sessions, not any reading or extraction of a user's actual codebase. That distinction is why Anthropic and its critics describe the same mechanism so differently.
Anthropic did not deny the code existed. Thariq Shihipar, an engineer on the Claude Code team, wrote on X that it was an experiment meant to curb account abuse by unauthorized resellers and to protect against distillation, the practice where a rival trains its own model on Claude's outputs to copy its capabilities more cheaply. He said stronger mitigations had since landed and that the team had already been planning to remove the mechanism.
Also read || How AI Is Fixing Medicine's Broken Information Gap
That tracking code was pulled out on July 1, the day after the Reddit disclosure and just before Alibaba's deadline. That means the mechanism had been running for roughly three months without being disclosed, which is the detail developers have focused on the most. No outside security firm has fully checked the claim yet, so it's still unclear whether this was a tool aimed only at specific bad actors or a broad filter that also caught ordinary developers in China who did nothing wrong.
That uncertainty is why reactions have been so split. Developers who found their sessions flagged see a company quietly profiling users by nationality. Anthropic's defenders see a reasonable, if clumsily disclosed, attempt to protect its product from being copied. Since distillation is a real and serious concern for any AI company, both readings can hold some truth, even as the wider US-China rivalry shapes how each side frames it.
Also read || NVIDIA vs Huawei: The Chip War Nobody Saw Coming
A dispute bigger than one ban
Alibaba's move doesn't exist in isolation. It follows a letter Anthropic sent to US senators on June 10, accusing operators linked to Alibaba's Qwen lab of running nearly 25,000 fraudulent accounts to extract Claude's software engineering and reasoning capabilities, generating around 28.8 million exchanges between April 22 and June 5. That figure reportedly exceeded three earlier distillation campaigns Anthropic had already flagged to Washington.
Alibaba has not responded publicly to those allegations of distillation. Anthropic already restricts Chinese companies and their foreign affiliates from using its models and has said it has been closing loopholes that let Chinese users reach Claude anyway. This policy tension likely shaped the disputed tracking code in the first place.
The wider pattern is worth noting too. As US model developers tighten access controls to prevent resale and distillation, Chinese firms have leaned harder on domestic alternatives. Alibaba already has its own Qwen models, and steering staff toward its in-house Qoder tool fits neatly into that same shift toward keeping developer tooling in-house rather than dependent on a foreign competitor it is simultaneously feuding with.
Claude Code has become one of Anthropic's fastest-growing products, widely used by developers to write and debug software from a command line rather than a chat window. That scale is exactly why a workplace-wide ban at a company the size of Alibaba matters. It signals that AI coding tools are now caught up in questions of sovereignty and enterprise trust, not just competition on raw capability.
Whether the code was targeted at specific bad actors or was a broader filter that swept up ordinary users remains unresolved, and Alibaba's internal notice doesn't settle that ambiguity either way. What's clear is the immediate outcome: Alibaba staff have until July 10 to drop Claude Code, Anthropic has removed the disputed code and explained its origin, and the underlying dispute over access, distillation, and trust between the two companies is far from over.
