Alibaba bans employees from using Anthropic tool

The Alibaba banned its employees from using Anthropic’s Claude Code programming tool, according to a source close to the company interviewed by Reuters.

The measure should come into force on July 10 and would have been taken after identifying that a version of Claude Code was capable of discreetly recognizing users located in China. As a result, the software was classified as high risk by the Chinese giant, which instructed employees to use their own programming tool.

It is worth remembering that Anthropic already restricts the use of its artificial intelligence models by Chinese companies and tries to block alternatives that allow developers based in China access to its technologies.

According to TechCrunch, a version of Claude Code was available through one of these alternatives, which led an Anthropic engineer, Thariq Shihipar, to speak out about the case in a post on the social network X.

Shihipar explained that the version of Claude Code capable of identifying users in China was part of “an experiment launched in March”, with the aim of preventing abuses committed by unauthorized resellers.

 

Hi, this is an experiment we launched in March that was meant to prevent account abuse from unauthorized resellers and protect against distillation.

The team has landed stronger mitigations since then and we’ve actually been meaning to take this down for a while. We merged the…

— Thariq (@trq212) June 30, 2026

US lifts restrictions on Anthropic’s most powerful models

Washington lifted restrictions that had forced Anthropic to block access to its two most advanced artificial intelligence models. The information was announced by the North American company itself, which began to reestablish access to the tools.

“We have received notification that the Department of Commerce has lifted export controls on Claude Fable 5 and Mythos 5,” Anthropic said on Tuesday, in a post on the social network X. “We will begin reestablishing access tomorrow,” it added.

On June 12, the United States government had abruptly ordered Anthropic to block access to the two cutting-edge models, alleging flaws in the security measures created to prevent the misuse of the tools in cyberattacks.

The forced withdrawal, considered unprecedented by a government, provoked criticism in several countries and reignited the debate about the digital sovereignty of nations dependent on North American technologies.

After intense negotiations in Washington, about which few details were released, access to Mythos 5 was released on Friday to a small group of “cyber defenders and infrastructure operators”, but only in the United States.

Foreign partners, including state cybersecurity agencies in Europe and Asia, remained without access at this stage. Anthropic, which has a turbulent relationship with the Trump administration, has not clarified whether the lifting of export controls will also allow these international partners to reintegrate.

The decision announced on Tuesday allows Fable 5 to become available again starting today. The model is a version aimed at the general public of Mythos, with restrictions related to cybersecurity and the risks of biological and chemical attacks.

OpenAI, Anthropic’s North American rival, on Friday launched its most powerful model yet, GPT-5.6, initially with limited access. For the first time, the company accepted that the United States government validates, client by client, which partners are authorized to use the technology.

For a long time opposed to the regulation of artificial intelligence, under the argument that rules could harm innovation in the face of competition with China, the Trump administration radically changed its position regarding the capabilities of the most advanced models.

Recent decisions in this area, taken in a legal scenario that is still unclear and controversial, indicate a resumption of control by the North American government over a technology considered strategic.

On Tuesday, CIA Director John Ratcliffe compared the most advanced models of artificial intelligence to “digital nuclear weapons” in a rare public demonstration in Washington.