Alibaba has instructed its employees to stop using Anthropic's Claude coding tool, according to a source with knowledge of the decision reported by Reuters. The move signals growing concern inside the Chinese tech giant about data security and competitive sensitivity when using AI tools developed by foreign rivals.
A Ban Rooted in Security and Competition
The decision reflects a broader pattern among large technology companies scrutinizing which AI tools their engineers can access. Alibaba, which operates its own large language model suite under the Qwen brand, has an obvious competitive interest in steering employees away from tools built by Anthropic. The company has not issued a public statement confirming the ban, and the full scope of restrictions remains unclear. What is clear is that the relationship between Alibaba and Anthropic has grown increasingly complicated over the past year.
Key Facts
- Alibaba has reportedly banned employees from using Claude Code, Anthropic's AI-powered coding assistant.
- The ban was reported by Reuters, citing a single source familiar with the matter.
- Alibaba has its own competing AI model family, Qwen, which includes coding-focused variants.
- Anthropic has previously accused Alibaba of misusing access to Claude through unauthorized channels.
- Security concerns around AI coding tools in enterprise settings have grown industry-wide in 2025.
Security concerns around AI coding assistants have intensified across the industry. Earlier this year, reporting surfaced that Microsoft flagged Claude's coding tool as a potential security risk, warning that it could inadvertently expose sensitive credentials or internal code. Whether Alibaba's ban stems from similar technical concerns, competitive strategy, or both is not yet known. Given the scale of Alibaba's engineering workforce, the practical impact of such a restriction could be significant.
The ban reflects a deepening distrust between Alibaba and Anthropic that goes beyond simple business competition.Reuters source familiar with the matter
A Troubled History Between the Two Companies
The ban does not emerge in a vacuum. Anthropic has made serious accusations against Alibaba in recent months. The San Francisco-based AI safety company accused Alibaba of illicitly accessing Claude through unauthorized means, alleging that the Chinese company used the access to gather data for training its own competing models. Separately, investigators found that Alibaba used around 25,000 fake accounts to extract data from Claude in what appeared to be a systematic effort to replicate the model's capabilities, a practice known as model distillation. Anthropic has taken legal and policy steps in response to those alleged violations.
Against that backdrop, Alibaba banning its own staff from using Claude Code reads as much like a defensive legal and reputational move as it does a straightforward security policy. The company would have a difficult time defending continued employee use of a tool made by a company that has accused it of theft and unauthorized access. From Alibaba's perspective, cutting off internal use may be the simplest way to limit further exposure.
For Anthropic, the ban represents a lost commercial opportunity but perhaps not a surprising one. Claude Code has been positioned as one of the flagship products in Claude's model family, and enterprise adoption has been a key part of the company's revenue strategy. Losing access inside one of the world's largest technology companies, even under these circumstances, is a meaningful setback in the competition for developer mindshare.
The broader picture is one of accelerating fragmentation in the global AI tools market. As geopolitical tensions between the United States and China continue to shape technology policy, decisions about which AI products employees can use are increasingly driven by factors well beyond simple capability comparisons. Alibaba's move may well encourage other Chinese firms to audit their own internal AI tool usage, particularly where tools made by U.S. companies are involved.
