Anthropic Rolls Back Claude ‘Spyware’ Code After Chinese AI User Tracking Sparks Backlash

Anthropic has rolled back controversial code embedded in its Claude Code assistant after reports revealed it was quietly collecting information that could identify users linked to Chinese AI companies, reigniting concerns over privacy, AI security and the escalating technology rivalry between the US and China.

The issue came to light after it was reported that Claude Code contained functionality that checked users’ time zones, proxy settings and possible affiliations with Chinese AI labs. Anthropic later acknowledged the feature, describing it as an internal experiment designed to combat account abuse and AI model distillation.

Anthropic Says Tracking Feature Was an ‘Experiment’

Responding to criticism on X, Thariq Shihipar, a member of Anthropic’s technical staff, said the feature was not intended for long-term deployment.

“An experiment we launched in March that was meant to prevent account abuse from unauthorized resellers and protect against distillation,” Shihipar wrote. He added that Anthropic had “landed stronger mitigations” and had “been meaning to take this down for a while,” noting that the code “should be fully rolled back” with the latest Claude Code release.

According to reports, the code examined Claude Code’s base URL environment variable used for API routing. If it detected a custom proxy, it performed additional checks, including the user’s system time zone and whether the hostname matched a list of known Chinese AI labs, AI companies, resellers or gateway domains.

Anthropic has repeatedly accused several Chinese AI firms, including DeepSeek, Moonshot AI, MiniMax and Alibaba, of using Claude to train competing AI models through a practice known as distillation.

The company previously alleged that entities affiliated with Alibaba created more than 25,000 fraudulent Claude accounts, generating over 28.8 million interactions between April 22 and June 5 in an effort to extract model capabilities.

The latest controversy comes amid growing geopolitical tensions over advanced AI systems. While the White House recently lifted restrictions on foreign access to Anthropic’s frontier models, businesses continue to express concerns about potential US control over advanced AI services through a national security “kill switch.”

The issue first gained wider attention after security researcher Alexander Hanf reported discovering the code on his Mac after installing Claude Desktop, despite not using any Anthropic browser extensions.

Critics argued that silently monitoring user attributes, even for security purposes, sets a troubling precedent.

One Reddit user warned, “Today it’s a timezone check. Tomorrow, it could be system sabotage or data exfiltration.”

Multiple media outlets said Anthropic referred questions about the feature to Shihipar’s X post, offering few additional details about how the experiment operated or what the company meant by its “stronger mitigations”.

The incident underscores how intensifying US-China competition is increasingly influencing AI product development and security decisions.

Anthropic CEO Dario Amodei has repeatedly warned policymakers about Chinese AI distillation efforts while also advocating international cooperation on frontier AI safety. Meanwhile, Sam Altman has called for a US-led global forum to establish international AI safety standards.

The Claude Code episode highlights the difficult balance AI companies now face between protecting proprietary models from misuse and maintaining transparency and user trust as geopolitical competition reshapes the industry.…Read more by Bhaswati Guha Majumder