Exposing Indian data to foreign LLMs risky, says expert

AS Lakshminarayan, Managing Director and CEO of Tata Communications, has warned that if India shares its data with foreign Large Language Models (LLMs), it risks losing control over its valuable knowledge, potentially leading to its commoditisation. He emphasised the necessity for India to develop its own AI models, whether they are large language models or smaller variants, to maintain authority over Indian data (information).

“India should retain control over its own data, which is sovereign property. If we expose our data to external LLMs, the knowledge becomes commoditised,” Lakshminarayan told Outlook Business.

American restrictions

Recently, the United States has tightened its export controls on artificial intelligence (AI) chips, and with the possibility of Trump returning to power, there are concerns that these restrictions may become even more stringent.

For context, in January 2025, the US government enacted limitations on the export of advanced AI chips to nations outside its primary allies, which include the UK, Canada, Germany, Japan, South Korea and Taiwan. Currently, countries not classified as close allies, such as India, are permitted to import up to 50,000 advanced AI chips without a license until 2027.

What Tata offers

Lakshminarayan made these remarks at the unveiling of Tata Communications’ next-generation cloud solution, Tata Communications Vayu, in Mumbai. The company asserts that this cloud offering is built on over 10 years of expertise in cloud computing, ranging from single-product to multi-product solutions. In essence, Cloud Fabric represents an advanced cloud infrastructure that integrates various cloud platforms and computing resources into a unified network.

Vayu AI Cloud offers on-demand access to powerful NVIDIA GPUs, facilitating the training and deployment of AI models at scale. In October 2024, the company partnered with chipmaker NVIDIA to incorporate Hopper GPUs, further enhancing its AI cloud infrastructure. To clarify, NVIDIA Hopper GPUs are high-performance graphics processing units specifically engineered for AI, data cecentresand supercomputing applications.

The cloud platform is versatile and caters to various industries. However, it has significant experience collaborating with government entities, as well as in the banking, financial services, insurance, manufacturing, and retail sectors, as noted by Bhaskar Gorti, Executive Vice President of Cloud and Cybersecurity Services at Tata Communications.

Gotri said, “Tata Communications is engaged with various government entities, including the National Health Authority (NHA) and state governments, to support cloud infrastructure for large-scale projects. However, details on state-level engagements cannot be disclosed at this time.”

In the third quarter of FY25, Tata Communications announced a net profit of Rs 236 crore, a significant rise from Rs 45 crore during the same period last year. The company's operating revenue also saw an increase, reaching Rs 5,798 crore, up from Rs 5,588 crore in the previous year's comparable quarter.