Many media reports indicate that Ali Baba will ban employees from using the Anthropic programming tool Claude Code from July 10. This adjustment appears at a time when Anthropic continues to tighten access to Chinese users, and reflects the tendency of large technology companies to promote their own in-house AI tools.
Ali is using his own tools.
The report mentions that Ali Baba has classified Claude Code as a high-risk software and has asked employees to switch to the company's own Qoder. The public information available does not indicate whether the ban covers all sectors and does not disclose more specific details of internal implementation.
Anthropic, tighten up the Chinese visit.
Anthropic had previously banned Chinese companies, as well as overseas entities held by those companies, from using its model. Recently, it has also been noted that the company is further closing the loophole that allows Chinese users to access Claude.
A recently circulated post on Reddit states that a version of Claude Code was able to identify Chinese users on the backstage. In response, Anthropic employee Tharik Shihipar stated on platform X that it was an experiment initiated by the team in March to prevent unauthorized distributors from abusing accounts and to reduce the risk of model distillation.
The old plan is offline.
Shihipar stated that the team had subsequently deployed more restrictive measures and that the original pilot programme had been ready for removal. Model distillation is usually used to train another model using the output of one model.
According to the current information, the Ali Baba operation was an internal tool management adjustment, but the events also showed that the differentiation around model access restrictions, account control and self-study substitution by enterprises was growing.