Alibaba has prohibited all employees from using Anthropic's Claude Code starting July 10, classifying the tool as high-risk software after security researchers discovered it contained hidden mechanisms to detect and track users based in China or affiliated with Chinese AI laboratories. The ban covers not just Claude Code but also Anthropic's Sonnet, Opus, and Fable language models, and represents a dramatic escalation in one of the AI industry's most contentious corporate disputes.

The discovery that Claude Code versions from April 2026 onward contained covert tracking features has sent shockwaves through developer communities and deepened the already fractious relationship between the two companies. According to Alibaba's internal notice, the tool was flagged as having "back-door risk" and classified as "high-risk software with security vulnerabilities". The company is directing its roughly 124,000 employees to migrate to Qoder, its proprietary in-house coding platform, which simultaneously launched an enterprise edition.

What Exactly Was Hidden Inside Claude Code?

Security researchers uncovered that Claude Code versions starting with 2.1.91, released in April 2026, contained embedded detection mechanisms that went far beyond standard telemetry. The tracking system worked by reading system time zones such as Asia/Shanghai or Asia/Urumqi, monitoring proxy servers and custom API addresses for keywords tied to major Chinese technology companies and AI labs, and using steganographic techniques to transmit user environmental data back to Anthropic.

The tool also altered how it processed system information in ways designed to evade detection. In the version described as starting with 2.1.91, the tool changed the date format in system prompts from "2026-06-30" to "2026/06/30" and replaced apostrophes with Unicode characters that are indistinguishable to the human eye. These obfuscation techniques suggest the tracking was intentionally concealed rather than an accidental data collection practice.

"An experiment launched in March aimed at preventing account abuse by unauthorized resellers and protecting against distillation," said Thariq Shihipar, an Anthropic engineer, regarding the tracking feature. He added that "the feature was slated for removal as stronger mitigations were subsequently introduced."

Thariq Shihipar, Engineer at Anthropic

Shihipar acknowledged on social media on July 2 that the tracking code existed and confirmed the code was rolled back in a July 1 update. However, the acknowledgment came only after security researchers posted their findings on platforms including Reddit and GitHub, suggesting the feature would have remained undisclosed without external discovery.

How Did This Dispute Between Anthropic and Alibaba Begin?

The Claude Code ban cannot be separated from a broader and more serious accusation that Anthropic leveled against Alibaba in June. Anthropic sent a letter to the U.S. Senate Banking Committee on June 10 accusing Alibaba's Qwen AI lab of conducting "industrial-scale model distillation," a technique where developers use a more advanced model's outputs to train cheaper, smaller models that approximate similar performance.

According to Anthropic's allegations, Qwen created approximately 25,000 fraudulent accounts between April 22 and June 5, 2026, generating roughly 28.8 million conversations to systematically extract knowledge from Claude's most capable models. Anthropic told lawmakers it is "the only frontier AI company that restricts service to Chinese entity-owned companies, even overseas subsidiaries".

Anthropic

The distillation allegations reveal how Chinese firms circumvented both U.S. export controls and Anthropic's terms of service. Ant Group and other Chinese companies reportedly accessed Claude through overseas affiliates in Singapore and via Microsoft Azure APIs, creating a workaround that Anthropic views as a coordinated effort to steal its intellectual property. In response, Anthropic implemented large-scale account restrictions on Chinese users in late June.

Steps Alibaba Is Taking to Enforce the Ban

• Mandatory Uninstallation: All employees must remove Claude Code and all Anthropic model series, including Sonnet, Opus, and Fable, from their workplace devices by July 10, 2026.
• Migration to Qoder: Alibaba is directing staff to switch to Qoder, its self-developed coding platform, which launched a new enterprise edition simultaneously with the ban announcement.
• Internal Classification: Claude Code has been added to Alibaba's list of high-risk software with security vulnerabilities, signaling to employees that use of the tool violates company policy.

The timing of Alibaba's migration to Qoder is strategically convenient for the company, which has invested heavily in its Qwen family of AI models and developer tools. By mandating Qoder internally, Alibaba ensures a large captive base of enterprise users whose usage data can further refine the product before external commercialization. This move also allows Alibaba to accelerate adoption of its own AI ecosystem while reducing dependence on Western AI tools.

What Does This Mean for the Broader AI Industry?

The Claude Code dispute sits within a broader deterioration of U.S.-China technology relations and represents a new front in AI competition where model-level intellectual property, not just chips and hardware, is becoming a flashpoint. Anthropic's lobbying for export controls or sanctions specifically targeting Qwen demonstrates how aggressively frontier AI companies are now willing to engage with U.S. policymakers to protect their models.

Anthropic explicitly prohibits Claude Code access from unsupported regions including China, but enforcement has proven difficult against determined circumvention. The episode highlights the fundamental challenge frontier AI companies face in controlling who uses their models, particularly when those models are deployed as developer tools with deep system access to local file systems and network configurations.

The discovery of the hidden tracking feature has also raised concerns among developers who granted Claude Code deep access to their systems. "Today it's a timezone check. Tomorrow, it could be system sabotage or data exfiltration," one Reddit user wrote, capturing the sentiment among developers who now question what other covert mechanisms might exist in widely used AI tools.

Alibaba's stock price reflected the tensions surrounding the dispute. Shares fell 1.9% on the day of the announcement, trading at $96.14 on the New York Stock Exchange and down more than 34% year-to-date. The company's market capitalization stands at approximately $231 billion, and shares remain near their 52-week low of $91.99, well below the 52-week high of $192.67.

Beyond the immediate corporate dispute, the Claude Code ban underscores how geopolitical tensions are reshaping the AI industry. As U.S. and Chinese companies compete for dominance in artificial intelligence, the tools developers use daily are becoming battlegrounds for control over model access, intellectual property protection, and national security concerns. The resolution of this conflict will likely shape how frontier AI companies approach security, transparency, and international access for years to come.