Alibaba bans Claude Code over alleged backdoor security concerns

Alibaba has banned employees from using Anthropic’s Claude Code in workplace environments from July 10 over alleged security concerns involving embedded backdoors, according to a person familiar with the decision.

According to a source familiar with the matter, the restriction will apply across Alibaba’s internal work environments and takes effect on July 10. The person said the company reached the decision because of alleged security risks linked to embedded backdoors in the coding assistant.

As of publication time, Alibaba has not issued an official statement, and no further details about the alleged security concerns or the scope of the restriction were disclosed.

Claude faces another enterprise setback

The latest development comes just weeks after Anthropic’s Claude models lost access to another major enterprise customer group in Hong Kong. In June, the Financial Times reported that JPMorgan had stopped employees in Hong Kong from selecting Claude models from the bank’s approved list of large language models because of Anthropic’s licensing terms governing where the models could be used.

The report said Goldman Sachs had previously introduced a similar restriction after determining that Anthropic’s terms of service excluded use across Greater China, including Hong Kong. Anthropic later told the Financial Times that Claude had never been officially supported in Hong Kong, while JPMorgan declined to comment.

Those restrictions added to concerns among some financial institutions in Hong Kong as advanced AI tools become more deeply integrated into software development, research, and financial services workflows, according to the Financial Times.

Anthropic recently restored its newest models

The Alibaba decision also follows a turbulent few weeks for Anthropic’s latest AI systems. On July 1, the company restored public access to its Claude Fable 5 and Mythos 5 models after U.S. authorities lifted export restrictions that had forced Anthropic to suspend them in June.

Anthropic said it resumed deployment after what it described as productive discussions with U.S. officials and added new classifiers designed to detect and block more cybersecurity-related tasks. The company said the additional safeguards addressed government concerns over possible misuse through jailbreak techniques.

While defending its technology, Anthropic argued that the reported jailbreak involved a limited method rather than a universal bypass of the models’ safety protections. The company also announced expanded cooperation with the U.S. government on model testing, safety evaluations, misuse tracking, and information sharing related to jailbreak risks.