Key insights
- Starting July 10, Alibaba will disable Claude Code in the workplace.
- The ban comes as part of mounting tensions between Alibaba and Anthropic regarding claims to AI model distillation.
- Chinese AI firms continue shifting toward domestic models as U.S. companies tighten access controls.
Alibaba will ban employees from using Claude Code across its workplace environments from July 10 after identifying alleged security risks linked to the AI coding assistant. The decision adds another layer to the growing dispute between Alibaba and Anthropic as both companies compete in an increasingly divided global artificial intelligence market.
The restriction also reflects broader concerns over AI security, export controls, and access to advanced models. It comes only weeks after Anthropic accused parties connected to Alibaba of attempting to extract capabilities from its Claude models through model distillation.
| Key detail | Information |
|---|---|
| Company imposing ban | Alibaba |
| AI tool | Claude Code |
| Effective date | July 10 |
| Reason cited | Alleged security concerns and embedded backdoor risks |
| Recommended alternative | Qoder |
| Source | Person familiar with the decision, Yicai and Reuters |
Security concerns intensify amid AI rivalry
Alibaba informed employees that Claude Code will no longer be permitted in internal workplace environments beginning July 10, according to a person familiar with the matter. Chinese newspaper Ming Pao, citing financial outlet Yicai, reported that Alibaba had classified the software as high risk and would recommend employees use its own coding platform, Qoder.
The source told Reuters the restriction applies across Alibaba’s workplace systems because of alleged security concerns involving embedded backdoors. However, Alibaba has not issued an official statement explaining the findings or providing technical details supporting the decision.
The latest action follows reports from developers that the coding assistant contained mechanisms capable of inspecting user environments. Those reports said the software checked information such as time zones and proxy settings before inserting subtle markers into prompts transmitted to Anthropic’s servers.
An Anthropic employee later said on X that the feature formed part of an experiment introduced in March. According to the employee, the company designed the mechanism to prevent abuse by unauthorized resellers while limiting attempts to distill its AI models.
Meanwhile, Hong Kong-listed Alibaba shares fell 0.7% during Thursday trading, while the broader Hang Seng Index gained 1.3%.
Distillation dispute expands beyond technology
Relations between Alibaba and Anthropic have deteriorated in recent weeks after Anthropic publicly accused operators linked to Alibaba and its Qwen AI division of conducting a large-scale model distillation campaign.
Anthropic told two U.S. senators in a letter reviewed by Reuters that model distillation could help China narrow the gap with its advanced Mythos Preview models. The company argued that unauthorized extraction of model capabilities threatens years of research and development.
Reuters also reported that Anthropic considers its China restrictions difficult to enforce for individual users. People can route internet traffic through servers located in the United States to disguise their location. Businesses, however, generally face greater legal and regulatory obligations, making compliance easier to enforce.
Alibaba has not publicly responded to Anthropic’s allegations. Neither company immediately commented on Reuters’ latest report regarding the workplace ban.
The latest restriction follows another setback for Anthropic in Hong Kong. In June, the Financial Times reported that JPMorgan removed Claude models from its approved AI model list for employees in Hong Kong because of licensing restrictions. Goldman Sachs reportedly implemented similar limitations after determining Anthropic’s terms excluded use across Greater China, including Hong Kong.
Competitive landscape continues to shift
The dispute arrives as Chinese AI companies increase investment in domestic alternatives. Alibaba continues expanding its Qwen family of large language models while competing against DeepSeek, Tencent, Baidu, Moonshot AI, and Zhipu AI.
At the same time, Anthropic recently restored public access to its Claude Fable 5 and Mythos 5 models after U.S. authorities lifted temporary export restrictions. The company also introduced additional safeguards designed to detect cybersecurity-related misuse and improve protections against jailbreak attempts.
The industry is continuing its shift towards more stringent regulations of advanced AI systems. U.S. developers are beefing up protections against unauthorized download, resale and model distillation, and Chinese developers are increasingly turning to local models to lessen reliance on foreign platforms.
- Compliance risks are now a factor in the choice of AI tools for companies.
- As governments grow more vigilant over advanced AI models, security has become a strategic concern in AI.
- AI vendors in the United States and China are increasingly vying for enterprise software business.
Conclusion
The move by Alibaba, which will ban Claude Code from workplaces, is another step up in the increasingly fierce technology wars between the Chinese and U.S. AI developers. This is not just about software policy, but also about heightened awareness of data security, compliance, and protection of sophisticated AI models.
Trust and regulatory clarity will likely be as important as capabilities for enterprise adoption of AI tools as regulations become more stringent, and companies increase the level of safeguards. The split between Alibaba and Anthropic is a good example of the geopolitical dynamics playing out in the AI space.
