US State Dept Warns Over Alleged China AI IP Theft by DeepSeek

The U.S. State Department issued a global warning alleging that Chinese AI companies, including DeepSeek, are stealing intellectual property from American AI laboratories. This diplomatic cable signals escalating tensions in the ongoing technological rivalry between the U.S. and China.

Core Allegations and Warning Details

The State Department's cable, seen by Reuters, warns about the risks associated with using AI models that have been 'distilled' from proprietary U.S. AI models. The document instructs diplomatic staff to raise concerns with foreign counterparts regarding this alleged activity.

Key points from the warning include:

• The Concern: Widespread efforts by Chinese companies to steal intellectual property from U.S. AI labs.
• The Mechanism: The use of 'distillation,' a process where smaller AI models are trained using the output of larger, more expensive models to lower training costs.
• The Risk: The resulting foreign models may appear comparable on certain benchmarks but fail to replicate the full performance of the original system. Furthermore, the cable alleges these campaigns 'deliberately strip security protocols' and undermine mechanisms ensuring ideological neutrality.

DeepSeek and China's AI Advancement

The timing of the warning coincides with DeepSeek's recent activities, which appear to underscore China's growing technological autonomy. DeepSeek, a Chinese startup, launched a preview of a new model adapted for Huawei chip technology. Furthermore, reports indicate that DeepSeek reportedly provided early access to this model to Chinese companies rather than American engineers.

Government Reactions and Context

• U.S. Government Action: The cable, dated Friday, was sent to diplomatic and consular posts globally, signaling the administration's seriousness regarding the issue.
• Chinese Response: The Chinese Embassy in Washington has previously dismissed similar accusations as "baseless allegations," while affirming Beijing's commitment to protecting intellectual property rights.
• Prior Warnings: This follows earlier warnings from OpenAI, which alerted U.S. lawmakers that DeepSeek was reportedly targeting major U.S. AI firms to replicate models for its own training.

The memo and subsequent cable are set to raise tensions in the long-running tech war between the two superpowers, potentially undermining any recent detente efforts.