Alibaba reportedly bans employees from using Claude Code
Original reporting by TechCrunch
Alibaba's recent directive banning its employees from using Anthropic’s programming tool, Claude Code, marks a significant development in the ongoing saga of AI technology access, intellectual property, and data security, particularly concerning Chinese entities. Effective July 10, the Chinese e-commerce giant has reportedly classified Claude Code as “high-risk software,” a designation that underscores growing concerns among major corporations regarding the provenance, control, and potential vulnerabilities of third-party AI tools. Employees are now instructed to exclusively utilize Alibaba’s proprietary alternative, Qoder, signaling a broader strategic pivot towards internal AI solutions and digital sovereignty within the company’s vast operations.
Closing Loopholes
This internal ban unfolds against Anthropic's own aggressive, and at times covert, efforts to prevent unauthorized access to its cutting-edge models within China. Anthropic already explicitly prohibits Chinese companies, and any foreign entities they own, from using its AI tools, and has been actively working to close various known loopholes. A recent Reddit post brought to light a particularly sophisticated tactic: an experimental version of Claude Code allegedly capable of secretly identifying Chinese users. Anthropic’s Thariq Shihipar confirmed this was a March experiment designed to combat account abuse by unauthorized resellers and prevent "distillation," a practice where AI models are illicitly trained on the outputs of other systems without permission. While Shihipar stated Anthropic has since implemented more robust mitigations and intended to retire the experimental feature, the revelation illuminates the complex, and often secretive, strategies employed by AI developers to control access and protect their intellectual property in an increasingly fragmented global technological landscape.
Alibaba’s prohibition on employees using Anthropic’s Claude Code, prompted by Anthropic’s own efforts to restrict access for Chinese entities, highlights a significant hardening of the global tech divide. This move, stemming from the reported classification of Claude Code as “high-risk” software — and possibly intensified by Anthropic’s acknowledged “experiment” in secretly identifying Chinese users — underscores escalating concerns over data sovereignty and digital trust. By mandating a shift to its internal Qoder tool, Alibaba exemplifies a broader strategy among Chinese firms: prioritizing domestic AI alternatives to mitigate perceived security risks and reduce reliance on foreign technology subject to geopolitical pressures.
A Deepening Divide
This incident transcends a single company’s policy shift; it foreshadows an increasingly balkanized global AI landscape. As major corporations and nations grow more wary of foreign-developed AI, concerns over intellectual property, potential espionage, and data integrity will invariably accelerate the adoption of localized solutions. This trend will compel global AI developers to navigate distinct, regionally compliant versions of their models, or establish separate operational footprints altogether. For users, this could lead to divergent access to advanced AI based on geopolitical alignment. For the industry, it signals a fragmented, competitive market, potentially hindering universal AI advancement and cross-border collaboration. Ultimately, Alibaba’s decision reinforces a trajectory toward distinct technological ecosystems, fundamentally shaping AI’s development and governance for decades.
Frequently asked questions
- Why did Alibaba ban Anthropic's Claude Code for its employees?
- Alibaba has prohibited its employees from using Anthropic's Claude Code, categorizing it as high-risk software. The company is directing its workforce to utilize its internal AI programming tool, Qoder, instead. This move aligns with broader corporate policies regarding external third-party software, particularly concerning data security, intellectual property, and compliance, especially when internal alternatives are available for sensitive development work.
- Why does Anthropic restrict access to its AI models like Claude Code in China?
- Anthropic maintains a policy of restricting access to its AI models, including Claude Code, for Chinese companies and foreign entities owned by them. This measure is implemented to enforce geographical restrictions, prevent unauthorized usage, and protect against practices like model distillation. Distillation involves training other AI models on Anthropic's outputs without permission, which the company actively works to prevent through various technical mitigations.
- Did Anthropic use a secret tool to identify unauthorized Chinese users of Claude Code?
- Anthropic did conduct an experiment involving a version of Claude Code capable of identifying Chinese users to combat unauthorized access and prevent model distillation. This experimental measure was part of efforts to protect against account abuse by unauthorized resellers. Anthropic has since implemented stronger and more permanent mitigations to manage access to its AI models, rendering the specific experimental tool no longer necessary.