DeepSeek is a Chinese Artificial Intelligence (AI) firm that made headlines earlier this year for developing an open-weight Large Language Model (LLM) that can rival GPT-4.
Compared to GPT-4, it achieved high efficiency at much lower costs using the Mixture of Experts (MoE) architecture. Also, the trained model weights (knowledge, in layman’s terms) are publicly released.
Nonetheless, GPT is proficient in more languages than its Chinese counterpart. Furthermore, public trust in GPT is higher because it has been around for longer and is continuously undergoing advancements.
However, US authorities have suggested that DeepSeek may be more than an AI Startup, and could be linked to China’s military-industrial complex.
US Believes DeepSeek Supported Chinese Military and Intelligence Operations
China and the US, two of the leading global superpowers, have been embroiled in a long trade war to increase their economic footprint. Washington believes that Beijing has an unfair advantage with DeepSeek.
A senior US State Department official believes that DeepSeek is feeding information to the Chinese military and intelligence apparatus, raising major security concerns. They stated:
“We understand that DeepSeek has willingly provided and will likely continue to provide support to China’s military and intelligence operations.”
US lawmakers had also previously raised the concern that DeepSeek was transmitting American users’ data to China, a serious allegation that was not addressed by the Chinese AI firm.
The official also hinted that Beijing may have tried to bypass America’s export restrictions to advance its military capabilities. The Hangzhou-based AI company has access to large volumes of US firm Nvidia’s high-end H100 chips.
China bypassed export controls to gain access to these chips. The official stated:
“DeepSeek sought to use shell companies in Southeast Asia to evade export controls, and DeepSeek is seeking to access data centres in Southeast Asia to remotely access U.S. chips.”
However, there was no clear indication that the firm had successfully evaded US export controls. Furthermore, DeepSeek didn’t issue an official response to the allegations.
Concerns About DeepSeek’s Legitimacy
LLMs like GPT and DeepSeek, with their conversational model, are designed for commercial usage. Knowledge has been made more accessible with these models, and personalised learning has become the norm.
If the allegations are proven to be true, DeepSeek has become a tool of the Chinese state, providing sensitive information beyond most commercial uses. This could also start a dangerous trend where states begin developing AI models solely for intelligence.
Most users are naturally oblivious to the breach of privacy, which raises questions about the legitimacy of DeepSeek and LLMs in general, too.
To manage such security concerns, official disclosures and conditions should be signed and verified before the individual uses DeepSeek. Furthermore, there is a desperate need for stricter checks and binding legislation on AI firms to oversee their ventures.
Stay tuned to Brandsynario for the latest news and updates