Unauthorized Access to Anthropic’s Claude AI Model Sold at Deep Discount in China

• May 10, 2026

A gray market has emerged in China offering discounted access to Anthropic’s Claude AI model at approximately 90% off the official price. A recent investigation by Zilan Qian, a researcher at the Oxford China Policy Lab, reveals how these proxy services operate using stolen credentials and unauthorized methods.

Gray Market Exploits Stolen Accounts and User Data for AI Model Refinement

According to the research, unauthorized providers in China resell access to Claude by leveraging compromised Anthropic user accounts. This practice allows them to offer the AI service at a fraction of the legitimate cost, undercutting official offerings and bypassing normal usage restrictions.

These sellers are not only exploiting account theft but also engaging in deceptive tactics such as substituting models. By rerouting users to different AI models, the gray market actors manipulate expected service delivery without users’ knowledge.

More notably, data collected from these illicit transactions—including user queries and AI-generated responses—are reportedly used for training and refining other AI models. This unpermitted usage raises significant concerns about data privacy and intellectual property rights within the AI ecosystem.

While official pricing and licensing policies govern legitimate access to Claude, the gray market circumvents those frameworks, providing cheaper but unsecure services. This parallel distribution channel poses challenges to AI providers, not only in protecting their proprietary technology from unauthorized use but also in safeguarding users’ personal and interaction data.

The findings highlight the vulnerability of AI service providers to credential theft and the broader risks such gray markets introduce, including potential misuse of sensitive data for AI distillation or other development efforts. As AI becomes increasingly integrated into diverse applications, protecting both systems and users from exploitation remains a critical concern.

Industry stakeholders continue to assess how to address these unauthorized access points through stronger authentication, usage monitoring, and international cooperation to clamp down on illicit AI service distribution networks.

A gray market in China resells discounted access to Anthropic’s Claude AI, exploiting stolen accounts and collecting user data for AI model distillation.