Sarvam co-founder and CEO Pratyush Kumar has said that the recent suspension of Anthropic’s Fable 5 and Mythos 5 models for foreign nationals highlights the growing importance of “sovereign AI” and serves as a wake-up call for countries and companies that rely heavily on externally controlled technologies.
Reacting to the development, Kumar argued that access to cutting-edge AI should not be mistaken for ownership. According to him, the incident underscores the risks associated with depending on technologies that remain subject to decisions made outside a country’s control.
“The Fable ban is a strong reminder of why sovereignty matters,” Kumar wrote, adding that Sarvam is focused on building India’s first truly consequential AI company.
Anthropic announced that it had been directed by the US government to suspend access to its Fable 5 and Mythos 5 models for all foreign nationals, including foreign employees working within the company. As a result, the company said it had to immediately disable access to the two models to comply with export control regulations. However, access to other Claude models remains unaffected.
Kumar said the episode carries important lessons for AI users, businesses, and nations. He warned that companies relying on AI systems with external control mechanisms must recognize that they remain vulnerable to policy changes and restrictions.
He also predicted that advanced frontier AI systems would become increasingly restricted, even as general-purpose AI technologies become more widely available. In this scenario, countries and organizations would need the ability to develop, deploy, and improve AI systems within their own boundaries—a concept he described as “Sovereign AI.”
Calling sovereign AI the foundation of Sarvam’s vision, Kumar highlighted the company’s progress across infrastructure, models, products, and deployment capabilities.
According to him, Sarvam has become the first Indian company to train sovereign AI models at scale, using around 3,400 Nvidia H100 GPUs. He also revealed that the company has brought India’s first Blackwell cluster online and is targeting tens of megawatts of AI compute capacity in the country by 2027.
On the model front, Kumar said Sarvam has developed “Sarvam 105B,” which he described as India’s first sovereign AI model built entirely from scratch. The company is now working toward even larger trillion-parameter-class systems and plans to release a coding-focused model in the near future.
He added that usage of Sarvam’s hosted AI models has tripled over the past three months, while the company is preparing to launch a production-grade inference platform aimed at banks, governments, enterprises, startups, and developers.
Kumar also said that Sarvam’s voice AI products are already powering millions of interactions every day, while its document intelligence solutions are witnessing rapid growth. The company is further expanding into enterprise AI agents and platforms that allow organizations to customize AI systems for their specific requirements.
Looking ahead, Kumar said Sarvam is attracting researchers with frontier AI expertise and has begun establishing a presence in San Francisco to strengthen links between India’s AI ambitions and the global ecosystem.
“India has the talent, economic scale, ambition, and policy support needed to emerge as a major force in AI innovation,” Kumar said, ending his remarks with a message aimed at potential recruits: “We are hiring.”
Recent Random Post:
