Mark RadarMARK RADAR
About
EN
Sign in

Alibaba Reportedly Bans Anthropic Products and Orders Claude Uninstalled

3 reports · First detected 2026-07-03 · Last active 2026-07-06

Tensions between Alibaba and U.S. artificial intelligence company Anthropic stem from the latter's allegations that Chinese companies may be extracting model capabilities through data distillation, as well as controversy over markers found in Claude Code's system prompts. Because the dispute involves corporate code, internal data and supply-chain security, it has been interpreted as a further escalation of U.S.-China AI competition into the realm of national security.

The latest reports say Alibaba has instructed employees to stop using and uninstall Anthropic products, including Claude and Claude Code, by July 10, citing the need to guard against potential backdoors and data leaks. The company is also promoting its internally developed coding tool Qoder. The reports disclosed no procurement costs or losses, while neither Alibaba nor Anthropic has yet provided a full public explanation.

All Coverage

3 original reports

The Backstory

The history behind this event
Alibaba Bans Claude Code, Shifts Staff to Qoder2026-07-27 · 1 reports · similarity 0.84

Anthropic restricts Chinese-owned companies from accessing Claude, while the United States and China are tightening controls around advanced AI models, chips and data. The issue has become more sensitive amid a dispute in which Anthropic accused actors linked to Alibaba’s Qwen lab of improperly extracting Claude’s capabilities, a claim Alibaba denies. For Chinese technology groups, replacing foreign coding agents with domestic systems reduces security, compliance and supply risks while bringing models, compute and developer workflows under internal control.

An internal directive reported on July 3 classified Claude Code as high-risk software and ordered employees to uninstall it, with the ban effective July 10, 2026. Staff were directed to use Alibaba’s Qoder instead. The move followed scrutiny of code allegedly capable of identifying China-linked users or AI labs. Separately, Anthropic accused operators tied to Qwen of using about 25,000 accounts for 28.8 million Claude interactions between April and June 2026 in a model-distillation campaign. Alibaba denied wrongdoing, and no financial terms were disclosed.

Mark Radar|MARK RADAR

If you search news on Google, you can set Mark Radar as a preferred source—our coverage will show up more often in your results. Set as preferred source on Google →

All times are in Taipei time (GMT+8)