Anthropic's hidden user-tracking system fractures trust with Chinese market

Anthropic faces a credibility crisis as the disclosure of a covert tracking system targeting Chinese users has prompted major companies to distance themselves from its flagship Claude model. The surveillance mechanism, deployed without transparency, stands in stark contrast to the AI safety company's public positioning around user privacy.

The tracker and Alibaba's response#

Last week, Alibaba moved decisively to protect its operations from potential legal exposure, banning employees from using Claude Code at work, according to a report in the South China Morning Post. An internal company memo characterized the discovery as particularly alarming: Claude Code now sits on Alibaba's roster of high-risk software flagged for security vulnerabilities and backdoor risks.

The tracker appears designed to identify and monitor users connected to leading AI development labs in China, raising questions about whether Anthropic deployed the system to defend its market position or for other purposes.

Business pressure and uncomfortable tradeoffs#

Anthropic's situation reflects deeper tensions in the AI industry. The company operates in a landscape where cheaper alternatives—including open-source Chinese models that rival some Western offerings in capability—are gaining traction among cost-conscious enterprises. Fortune 500 executives are actively seeking lower-cost AI solutions, intensifying competitive pressure on Anthropic.

For Alibaba, the risks of non-compliance carry teeth. While individual users might sidestep location blockers through affordable privacy tools without serious consequences, Alibaba faces genuine legal and compliance exposure if discovered violating Anthropic's terms of service. That calculus shifted the decision from inconvenient to unacceptable.

The credibility gap#

Anthropic's challenge extends beyond lost contracts. The company built its brand partly on principled stances around AI safety and user rights. A covert surveillance system—regardless of its intended target—undercuts that positioning and may force difficult conversations with other users and partners about what security assurances from Anthropic actually mean. User trust in the AI sector remains fragile, and moves that feel deceptive tend to ripple broadly.