DeepSeek, Alibaba among major Chinese AI firms accused of stealing trade secrets of American counterparts

The United States on Tuesday accused Chinese Artificial Intelligence firms, including DeepSeek and Alibaba, of stealing trade secrets of their American counterparts to cut development costs and swiftly release coveted technologies to the market.As per a joint advisory by the Federal Bureau of Investigation, National Security Agency, and the Cybersecurity and Infrastructure Security Agency, major Chinese technology firms are extracting “restricted proprietary functionalities and capabilities of US frontier AI models", using "distillation" techniques, where smaller models are trained on more advanced architectures.The disclosures come as President Donald Trump prepares to discuss AI with Chinese leader Xi Jinping during his Washington visit, later this month.US Authorities and tech executives acknowledged that distillation can serve as a legitimate research method, but argued that the scale and targeted nature of Chinese efforts amounted to theft.The advisory stated that the scope and sophistication of these campaigns showed distillation was not merely supplementary but "the critical core" of these firms' AI development.Additionally, the agencies claimed that Chinese AI companies rely on a network of intermediaries, described as a "grey market of proxies known as 'transfer stations'", to bypass geographic limitations and safety measures put in place by US firms, with Alibaba Group and DeepSeek among the companies named as having run "high-volume knowledge distillation campaigns".US Treasury Secretary Scott Bessent cautioned that penalties could follow, writing on X that sanctions and Entity List designations would be considered against Chinese firms found to have crossed "into IP theft" through covert, large-scale distillation.Separately, Anthropic reported that three Chinese AI laboratories had run an "industrial-scale campaign" to extract capabilities from its Claude model, generating over 16 million exchanges through roughly 24,000 fraudulent accounts.Adding to these concerns, a Google Threat Intelligence report published the same day found hackers increasingly targeting AI assets, warning that proprietary models and source code have become prime targets for espionage, extortion and resource theft.Catch the latest World News and Live updates. Download the TOI app.