Anthropic Restrictions Put India’s AI Strategy in Focus
Anthropic Restrictions Put India’s AI Strategy in Focus
Anthropic has suspended access to its newly launched Fable 5 and Mythos 5 models for foreign nationals following a directive from the U.S. government. According to the company, the restriction also applies to its own employees who are not U.S. citizens. In India, the decision has reopened a larger debate about the country’s long-term AI strategy.
The issue is not limited to one company or one model release. For Indian founders, investors, enterprise technology leaders and policy experts, the move raises a strategic question: how much should one of the world’s largest AI markets depend on systems developed and governed elsewhere?
The timing is notable. Anthropic had recently announced a partnership with Tata Consultancy Services to support enterprise AI adoption in India. That agreement underlined the importance of India as a commercial market for U.S. frontier AI providers, but the access suspension has also highlighted the limits of that dependence.
A major market with limited control
India has become a key growth market for leading AI companies. Anthropic and OpenAI have both described the country as their second-largest market after the United States. Both companies have expanded their local presence through offices, hiring, partnerships and enterprise programs.
That growing footprint makes the current restriction more significant for Indian technology companies. If access to advanced AI models can change because of government decisions abroad, companies using those models face risks that are not purely technical. Product roadmaps, automation plans, customer deployments and cross-border engineering teams may all be affected.
Aakrit Vaish, founder of the Indian AI venture platform Activate, said the episode strengthens the argument for sovereign AI capabilities in India. He expects startups to look more seriously at open-source models and to reduce reliance on a small number of frontier model providers.
Startup competitiveness enters the debate
For startups operating across borders, the concern is especially practical. Vijay Rayapati, co-founder and CEO of Atomicwork, warned that companies with internationally distributed AI teams could face disadvantages if access to advanced systems increasingly depends on nationality or geopolitics.
Atomicwork has employees in the United States, while much of its product engineering team is based in Bengaluru. For companies with similar structures, unequal access to frontier models could influence development speed, product quality and competitive positioning.
The debate comes as parts of India’s technology sector are already assessing how AI could reshape global work. Opendoor, a U.S. real estate technology company, closed its India office less than two years after expanding in the country. Its CEO pointed to a focus on bringing operational work closer to U.S. customers and building smaller AI-native teams. The company did not quantify how much AI efficiency influenced the decision.
Open source and sovereign infrastructure
The Anthropic case has also led Indian technology leaders to revisit the role of open-source and smaller AI models. Zoho founder Sridhar Vembu argued that Indian organizations should adopt more open and compact models, including options developed in India and China.
Investor and former Infosys executive Mohandas Pai called for a much larger national AI program, including more investment in AI, compute, hardware and deep technology. He proposed an annual 500 billion rupee fund, roughly 5 billion U.S. dollars, and a 2 trillion rupee credit guarantee program, about 21 billion U.S. dollars, to support cloud infrastructure, hardware and semiconductor development.
Those figures would be far larger than India’s current central AI program. In 2024, New Delhi approved the IndiaAI Mission with funding of 103.72 billion rupees, about 1.2 billion U.S. dollars, over five years. The program is intended to expand compute capacity, support startups and encourage domestic AI development.
Not everyone sees funding as the main bottleneck. Lightspeed partner Hemant Mohapatra has argued that talent, access to compute and execution are the more decisive constraints. He estimated that training a frontier model can cost from hundreds of millions to several billion U.S. dollars, depending on the strategy.
India’s AI sector is still application-led
India has a large developer base and a fast-growing enterprise market for AI tools, but it remains a smaller player in frontier model development. Only a limited number of startups are building foundational AI models. Sarvam has released open-source models, while Krutrim has shifted toward cloud and AI infrastructure services after initially positioning itself around foundation model development.
Much of the country’s AI activity is focused on applications and specialized models built on top of existing foundation models. Avataar AI, for example, introduced a video-generation model positioned as a lower-cost alternative to products from Google, Kling, Luma and Runway.
Geopolitics becomes an enterprise AI risk
For business leaders, the broader lesson is that AI strategy is becoming inseparable from geopolitical risk. Access to models, compute and infrastructure can be shaped by security decisions, export rules and diplomatic priorities. That matters for startups, IT service providers, consultancies and enterprises integrating AI into core workflows.
Technology policy expert Prasanto Roy said the episode is likely to reinforce concerns in India about strategic autonomy. He compared the issue to earlier lessons countries drew from Russia losing access to parts of the global financial system after its invasion of Ukraine.
For India, the Anthropic episode is therefore less a short-term access problem than a stress test for AI dependence. It has intensified questions about model portability, open-source strategies, domestic compute capacity and the balance between global innovation and national control.