From Kiplinger · · 3m
China's AI theft threatens US tech dominance
Chinese companies are systematically stealing AI secrets from American firms, undermining US leadership in the technology race.
In brief
Chinese companies are systematically stealing AI secrets from American firms, undermining US leadership in the technology race. China is rapidly advancing its AI capabilities not through innovation but by stealing proprietary technology from US companies like OpenAI and Google, a national security threat that…
China's AI theft operation exposed
US cybersecurity agencies report that Chinese AI companies including DeepSeek, MoonshotAI, and Alibaba are systematically stealing secrets from American firms like OpenAI, Google, and Anthropic. The theft amounts to billions of dollars in development costs.
How the attacks work
Chinese companies use prompt injections, clever wording that tricks AI chatbots into revealing their secrets and bypassing safety guardrails. They also deploy vast query networks and use proxies to hide their tracks.
A simple but devastating weakness
Attackers simply ask AI systems to reveal their reasoning process step-by-step. This basic technique, called distillation, breaches terms of service and extracts capabilities worth billions in research and development.
No perfect defense yet
US AI companies are improving detection and customer verification, but attackers keep getting craftier. Agencies like CISA are pushing for coordinated ecosystem-wide responses, but the problem is likely to persist.
The national security stakes
Stolen AI technology could give China an unfair advantage in the global AI race. The US government views beating China in AI as a national security priority, since the global leader will control an incredibly powerful technology.
Dangerous AI could spread widely
Stolen AI models lack the safety guardrails of legitimate tools. This could enable criminals to develop bioweapons, build advanced military hardware, and launch cyberattacks. Authoritarian governments could deploy offensive cyber weapons and mass surveillance.
“Dangerous capabilities may proliferate with many protections stripped out.”
DeepSeek's cost claims are misleading
DeepSeek claimed it built advanced AI for just 5.6 million dollars. But the CISA report reveals this ignores the true cost of data acquired through extensive malicious distillation attacks on US companies.
China's innovation shortcut
The distillation attacks prove China hasn't found a cheap way to build advanced AI. Instead, Chinese advances come partly from siphoning US innovation, while still requiring huge spending on chips and data centers.
Cheaper models intensify competition
Chinese AI models like DeepSeek are gaining ground in the US market through lower prices. This competitive threat underscores the urgency of stopping distillation attacks before Chinese companies capture market share.
What investors should watch
The prevalence of distillation attacks reveals that China's AI advances rely on theft rather than innovation. Investors should monitor whether US companies can stop these attacks and maintain their technological edge.
The short version
China is rapidly advancing its AI capabilities not through innovation but by stealing proprietary technology from US companies like OpenAI and Google, a national security threat that companies and agencies are struggling to stop.
