India pushes its sovereign AI after US blocks Anthropic

What happened?

On June 12, 2025, the United States government ordered Anthropic to shut down access to its Fable 5 and Mythos 5 models as part of an export control directive restricting foreign citizens from using the country's most advanced AI systems. The measure directly affected India, Anthropic's second-largest market, where developers and companies relied on these models for their applications. According to The Next Web, the order was issued by the Bureau of Industry and Security (BIS) of the Department of Commerce, which updated restrictions on the export of AI technology, including models with advanced reasoning capabilities. Anthropic, based in San Francisco, had deployed Fable 5 and Mythos 5 in its data centers in the United States, but the new regulation prohibits non-US citizens from accessing these models from outside the country, even through APIs. This has left thousands of Indian developers without access to tools they used for tasks such as code generation, data analysis, and chatbots.

Historical context

This is not an isolated incident. In recent years, the United States has tightened restrictions on the export of critical technology, first with semiconductor chips and now with artificial intelligence models. The trade war with China and growing concerns about national security have led Washington to control access to its most sensitive technological assets. In October 2022, the BIS imposed restrictions on the export of advanced Nvidia and AMD chips to China, and in 2023 expanded these restrictions to semiconductor manufacturing equipment. In 2024, the Biden administration issued an executive order granting the Department of Commerce authority to regulate frontier AI models, citing national security risks. The June 12 order is the first direct application of this authority against a model provider. India, which has sought to position itself as a global tech hub, finds itself in an awkward position: on one hand, it depends on US technology; on the other, it seeks to reduce that dependence. The country has been a close partner of the United States in the Indo-Pacific, but this measure shows that cooperation does not exempt India from restrictions.

Impact on India

The shutdown of Anthropic's models has been a wake-up call for the Indian AI ecosystem. According to industry data, more than 5,000 Indian startups used Anthropic's APIs, and many had integrated Fable 5 into their workflows. Companies like Razorpay, Zomato, and Ola had begun experimenting with Mythos 5 to optimize their operations. Universities such as the Indian Institute of Technology (IIT) Bombay and the Indian Institute of Science (IISc) also used these models for research. The sudden disconnection has disrupted ongoing projects and forced companies to seek alternatives, such as Meta's open-source models (LLaMA 3) or local platforms like Sarvam AI and CoRover.ai. This has strengthened the argument of sovereign AI advocates: the need to develop indigenous infrastructure, from chips to foundational models, to avoid being at the mercy of external political decisions. The Indian government, through its 'IndiaAI' initiative, has allocated 10 billion rupees (approximately $120 million) to build a GPU cluster with 10,000 units, but the project is still in its early stages. Dependence on foreign chips remains a bottleneck, as India lacks advanced manufacturing capacity.

Comparison with past events

The situation recalls the Huawei ban in 2019, when the United States prevented the Chinese company from accessing US technology and services, including Android and Qualcomm chips. At that time, China accelerated its own technological development efforts, pushing companies like Huawei to create their own operating system (HarmonyOS) and invest in semiconductors. India appears to be following a similar path, albeit with an ecosystem more dependent on the West. Another relevant comparison is the 2023 restriction that prevented Nvidia from selling A100 and H100 chips to China, leading Chinese companies to develop alternatives like Huawei's chips (Ascend) and optimize the use of existing hardware. In India's case, the response has been more cautious: the government has convened experts to assess the impact and initiated dialogues with Anthropic to seek exemptions, but so far without results. Unlike China, India lacks a chip manufacturing ecosystem and large capital reserves for massive investments, limiting its ability to respond immediately.

Consequences and outlook

The Indian government is expected to increase investment in its 'IndiaAI' initiative, which aims to create indigenous language models and computing capabilities. Companies like Reliance Industries and Tata Group have already announced plans to develop AI infrastructure. Reliance, through its Jio Platforms division, has invested in AI startups and is building a data center with 1,000 MW capacity. Tata Group, meanwhile, has launched a language model called 'Tata Neu' for enterprise applications. However, the road is long: India lacks chip manufacturing capacity and the large GPU clusters that the US and China possess. According to a report by the India Semiconductor Industry Association, the country produces only 0.5% of the world's semiconductors and relies on imports for 95% of its needs. For readers, the lesson is clear: technological dependence carries geopolitical risks. Companies and developers should diversify their suppliers and consider open-source platforms or non-US alternatives. Additionally, the measure could accelerate the adoption of local AI models and foster collaboration between the public and private sectors in India. The movement for sovereign AI in India has gained a compelling argument. It remains to be seen whether the country can turn this crisis into an opportunity to achieve true technological autonomy, or whether restrictions will continue to hinder its growth in the field of artificial intelligence.