Indian artificial intelligence startup Sarvam AI has faced criticism online after introducing Chinese-developed AI models on its newly launched Token Factory platform, raising questions about its positioning as a company focused on building sovereign AI capabilities for India.
The Bengaluru-based startup unveiled Token Factory this week as an AI inference platform that allows developers and enterprises to access multiple AI models through a single application programming interface (API).
The platform offers Sarvam's proprietary model alongside open-weight models developed by Chinese AI companies, including DeepSeek, Moonshot AI and Z.AI.
The inclusion of these models has prompted criticism on social media, particularly because Sarvam has been among the companies advocating greater technological independence and the development of domestic AI infrastructure.
Sarvam has received support under the government's IndiaAI Mission and was among the early Indian startups granted access to subsidised computing resources for developing advanced AI models.
The company has positioned itself around building AI systems tailored to India's languages, cultural contexts and domestic requirements, making the decision to offer foreign-developed models a subject of debate.
Responding to the criticism, Sarvam co-founder Pratyush Kumar maintained that the company's commitment to developing AI models for India remains unchanged.
Kumar said Sarvam intends to operate as a full-stack AI company, combining its own model development capabilities with commercial services that provide access to other AI systems.
He explained that selling AI tokens creates an additional revenue stream, which can support the company's operations as training and developing advanced AI models requires substantial investment.
According to Sarvam, its platform already processes more than 450 billion tokens daily using DeepSeek's V4.1 Flash model.
The company also clarified that the infrastructure supporting Token Factory is separate from the computing resources provided under the IndiaAI Mission.
Co-founder Vivek Raghavan said the chips used for the commercial platform were financed entirely through private investment rather than government subsidies.
The distinction comes amid questions about whether publicly supported AI infrastructure should be used to provide access to foreign-developed models.
Sarvam has also highlighted data residency as a central feature of the service. According to Raghavan, the platform hosts the models on infrastructure located in India, allowing customer data to remain within the country rather than being processed on overseas servers.
He argued that using AI services hosted outside India could expose users to additional data-related risks, particularly when information is transmitted to external infrastructure.
However, domestic hosting does not change the origin of the underlying AI models, which remains a separate consideration in discussions about technological independence.
Sarvam, backed by investors including Lightspeed and Khosla Ventures, was valued at approximately $1.5 billion in April 2026.
The controversy highlights questions surrounding India's AI development strategy, particularly the distinction between building domestic foundation models, operating locally controlled infrastructure and offering access to AI technologies developed internationally.
Disclaimer: This article may include information derived from interviews, press releases, public statements, research, company communications and other publicly available or third-party sources. Such material may be summarised, paraphrased or contextualised for journalistic and editorial purposes. All rights in third-party content remain with their respective owners.