Union Minister: DeepSeek On Indian Servers Will Soon Address Privacy Issues

Nitin Konde, Varun Bhardwaj

01 Feb 2025 — 2 min read

DeepSeek to Operate on Indian Servers, Says Union Minister

India will soon host the Chinese AI platform DeepSeek on local servers, according to Union IT Minister Ashwini Vaishnaw, in order to allay privacy worries about cross-border data transfers. Since DeepSeek is an open-source AI model that may be hosted on Indian servers, the minister stated as much during a news conference on the IndiaAI Mission. "This project is currently being worked on by our team. The framework and other information, including the number of servers and capacity needed, have been prepared, Vaishnaw stated.

Strengthening Country’s Data Protection System

In keeping with the nation's worries about data protection, Vaishnaw also stated that all of the open-source models will soon be housed on Indian servers. The IT community was recently rocked by DeepSeek's ability to create its own LLM using just 2,000 Nvidia H800 GPUs. For comparison, OpenAI's ChatGPT-4 uses one lakh GPUs, specifically the more sophisticated H100. Furthermore, the Chinese business asserted that, in contrast to OpenAI's $100 million investment, it was able to accomplish this with a mere $6 million.

The development coincides with the government's efforts to safeguard Indians' digital data, as data protection has become a major area of attention. The draft regulations for the Digital Personal Data Protection Act were recently made available for public comment until February 18 by the Ministry of Electronics and Information Technology (MeitY).

India’s Plan on Building its Own AI Foundation Model

According to the draft, data fiduciaries—entities that decide how personal data is processed—must give data principals—end users whose data they are collecting—all the specific information they need to give them so they can decide whether or not to use their personal data. In the meantime, DeepSeek's success has sparked a discussion in India about the necessity of creating its own core models. Vaishnaw stated during the conference that India intends to use the IndiaAI Mission to develop its own huge language models.

Abhishek Singh, additional secretary, Union Ministry of Electronics and Information Technology (MeitY), stated during his visit to GIFT City in Gujarat on 27 January that India is seeking proposals for developing its own foundational model in order to compete with foreign AI foundational models that serve as the basis for numerous applications such as ChatGPT and Gemini. He added that how India creates a basic model has been a major problem.

At the moment, every core model that a nation uses, including ChatGPT, Llama 3, Gemini, and Claude, is foreign. China has now developed Deepseek and other models as well. To create an Indian foundational model, India must invest and offer financial assistance. The foreign fundamental models can provide incorrect or unsuitable answers because they were trained on western data sets and are not adapted to Indian languages and contexts.