Regulatory Scrutiny: Beijing Investigates AI Giants Over Alleged Data Transfers to Anthropic
The landscape of Chinese artificial intelligence is facing a significant shake-up as the Cyberspace Administration of China (CAC) launches a formal inquiry into two of the nation’s most promising AI startups: DeepSeek and Moonshot AI. This regulatory intervention follows allegations that these firms may have funneled sensitive domestic data into the infrastructure of the U.S.-based AI powerhouse, Anthropic.
The Core Allegations
The investigation stems from a September 10 report published by Anthropic, which detailed unauthorized access to its Claude platform. According to the findings, the scale of the activity was substantial:
* Moonshot AI: Allegedly utilized over 5,380 deceptive accounts to facilitate more than 23 million interactions with Claude.
* DeepSeek: Reportedly generated upwards of 12.1 million exchanges within a concentrated two-week period this past July.
Initially, the CAC cast a wide net, summoning seven different companies mentioned in the Anthropic report. However, the focus has since sharpened specifically on DeepSeek and Moonshot, with regulators conducting on-site interviews with leadership and technical staff to determine the extent of the data exposure.
High-Stakes Timing
The timing of this probe could not be more precarious for either organization. DeepSeek is currently navigating the global stage, preparing to brief the United Nations Security Council on the inherent risks associated with advanced AI development. Simultaneously, Moonshot AI is in the midst of a sensitive financial transition, having recently submitted a confidential filing for a $3 billion initial public offering (IPO) in Hong Kong.
For these companies, the investigation represents more than just a technical audit; it is a test of their compliance with China’s stringent data security laws. Regulators are particularly concerned with whether proprietary information-ranging from state-linked corporate intelligence to sensitive military and law enforcement data-was inadvertently or intentionally transmitted to American servers.
The Broader Context of AI Sovereignty
This development highlights the growing tension between the rapid advancement of Large Language Models (LLMs) and the geopolitical boundaries of data governance. As AI labs globally race to improve their models, the “scraping” of data has become a contentious issue. While companies often use public data to train models, the transfer of potentially classified or sensitive information across borders triggers national security alarms.
In an era where AI models are increasingly viewed as strategic national assets, the ability to secure data pipelines is paramount. As the CAC continues its probe, the industry is watching closely to see if this marks a shift toward tighter state oversight of how Chinese AI labs interact with foreign platforms.
Myriad: Who will control Congress after the midterms? Click to make your prediction.
