India's AI Dependency Crisis: When Washington Pulls the Plug, Bengaluru Feels It First

When Anthropic pulled its most powerful AI models from broad access earlier this month under a US government export control directive, the shockwaves were felt around the world. But nowhere did the reaction land harder than in India.

India is Anthropic's second-largest market after the United States. It is also OpenAI's second-largest market. The country has more than 207,000 recognised startups, a rapidly expanding developer base, and an AI adoption curve that some analysts describe as the fastest outside the US and China. And almost all of it is built on a handful of foreign frontier AI models.

That dependence just became a national conversation.

What Happened

The US Commerce Department issued an export control directive requiring Anthropic to suspend access to its Fable 5 and Mythos models — its most capable systems — citing advanced cybersecurity capabilities that raised concerns in Washington. The move was not targeted at India specifically, but India felt it acutely because of how deeply its startup ecosystem had integrated Anthropic's models into production systems.

Within hours, Indian founders and investors were reacting publicly. Aakrit Vaish, founder of Indian AI venture platform Activate, said he woke up "shocked and confused" by the announcement. He told TechCrunch the episode "completely changes things" and that he would be encouraging portfolio companies to reduce their dependence on a small number of frontier AI providers.

Vijay Rayapati, co-founder and CEO of Atomicwork — a startup with around 25 employees in the US and a large product engineering team in Bengaluru — described the risk more starkly. If access to advanced AI increasingly becomes subject to geopolitical restrictions, he said, startups with teams spanning multiple countries face an entirely new category of business risk.

The Voices Calling for a Rethink

The most prominent intervention came from Sridhar Vembu, founder of Zoho, who wrote on X that the episode proved "technology is the ultimate weapon." His prescription was direct: Indian organisations should embrace smaller models, including Indian and Chinese open-source alternatives, and reduce reliance on a handful of US frontier providers.

Mohandas Pai, former Infosys executive and prominent investor, went further. He called for a national AI mission and urged the Indian government to create an annual ₹500 billion — roughly $5 billion — fund for AI and deep tech. "We are way behind and need a national mission to get going quickly," Pai wrote.

These are not fringe voices. Vembu and Pai are among the most respected figures in India's technology establishment. The fact that both responded to a single foreign policy decision with calls for strategic autonomy signals how seriously the episode was taken.

The Structural Problem

India's AI challenge has two layers. The first is access — the risk that geopolitical decisions in Washington or elsewhere can suddenly cut off Indian developers and startups from the tools they depend on. The second is capability — the question of whether India can build or cultivate the domestic AI infrastructure to avoid that dependency in the first place.

On access: India's position as a massive consumer of frontier AI makes it a valued customer, not a strategic partner. When the US government decides to restrict a model for national security reasons, India has no seat at that table. It finds out when everyone else does.

On capability: India has genuine strengths — a large developer talent pool, strong mathematical foundations, and a growing number of AI research labs. But it lacks the compute infrastructure, capital concentration, and policy coordination to compete with the US or China at the frontier level in the near term. The gap is real.

What Is Being Done

The Indian government has not been entirely passive. The IndiaAI mission, announced in 2024, has allocated funds for compute infrastructure and is supporting indigenous model development. Several Indian startups — including Sarvam AI — have been building smaller, India-focused language models optimised for Indian languages and use cases.

But these efforts, while meaningful, are not yet at a scale that could substitute for frontier models in enterprise and developer workflows. The honest assessment is that India's AI ecosystem will remain significantly dependent on foreign frontier models for the foreseeable future — which makes the access question urgent.

The Wider Lesson

The Anthropic episode is a preview of a dynamic that will only intensify as AI models become more capable and more strategically important. The most powerful AI systems are being built in the US, and the US government is increasingly willing to treat them as strategic assets — controlling who can access them and under what conditions.

For India, that is not just a technology policy question. It is a question about economic sovereignty in an era where intelligence infrastructure may matter as much as physical infrastructure. The country that controls the most capable AI models has leverage over every other country that depends on them.

The debate that Anthropic's model restrictions triggered in India will not go away. It is, if anything, just getting started.

Published June 27, 2026.