Anthropic Limits New AI Model, India Highlights Technology Dependence

JAKARTA - Anthropic's decision to restrict access to its latest AI model has India questioning its reliance on technology built and controlled by other countries.

As reported by TechCrunch, citing Monday, June 15, Anthropic received a directive from the United States government to stop access to the Fable 5 and Mythos 5 models for all foreign citizens, including the company's own foreign employees.

The decision comes shortly after Anthropic entered into a partnership with Tata Consultancy Services or TCS, an Indian information technology services giant, to expand the use of AI in the Indian business sector.

Anthropic denied the US government's assessment and stated that the restrictions should not have been imposed. However, this case sparked a debate among India's startup founders, investors, and policy experts.

The question is whether India should accelerate the development of domestic AI, increase the use of open source models, or remain dependent on a handful of US-based AI frontier providers? Frontier AI is the most advanced artificial intelligence model that is at the forefront of current technological developments.

India is an important market for global AI companies. Anthropic and OpenAI both call India their second largest market after the United States. Both have opened offices, recruited local staff, forged partnerships, and targeted Indian developers, startups, and companies.

Aakrit Vaish, founder of the AI platform Activate India, called Anthropic's decision a shift in the way India views AI sovereignty.

"This really changes everything," Vaish said, as quoted by TechCrunch.

He said the incident strengthened India's reason for building its own AI capabilities. Vaish also predicted that more startups would use open source models, that is, models whose code can be accessed, studied, and developed again by the public or a certain community.

In an interview with TechCrunch, Vijay Rayapati, co-founder and CEO of Atomicwork, said that restricting AI access could affect the competitiveness of startups with cross-country teams. Atomicwork has about 25 employees in the US, but most of its product teams are in Bengaluru, India.

"If your AI team is not entirely made up of US citizens, you are in a less competitive position," said Rayapati.

This debate comes as India's tech sector is also facing other pressures. Opendoor, an American property technology company, closed its office in India less than two years after expanding. CEO Kaz Nejatian said the company wanted to bring operational work closer to US customers and move to a smaller team based on AI.

The Anthropic case also sparked comments from Indian technology figures. Sridhar Vembu, founder of Zoho, wrote on X that "technology is the main weapon". He encouraged Indian organizations to use small and open source AI models, including from India and China.

Investor and former Infosys executive Mohandas Pai believes India needs a much bigger national AI strategy. He proposed an annual fund of 500 billion rupees or about 5 billion US dollars for AI and deep technology, as well as a 2 trillion rupee credit guarantee for cloud computing infrastructure, hardware, and semiconductors.

The proposal is much larger than the IndiaAI Mission approved by New Delhi in 2024, with a budget of 103.72 billion rupees or about 1.2 billion US dollars over five years.

Despite the increasing interest in AI, India is still relatively small in the development of basic models. Basic models are large AI models that form the foundation of various applications. Only a few startups are moving in that field, including Sarvam. Most of the Indian AI ecosystem is still focused on specific applications or models built on existing basic models.

However, not everyone sees capital as the main problem. Hemant Mohapatra, a partner at Lightspeed, assessed that the biggest challenge is talent, computing access, and execution capabilities. Training cutting-edge AI models can cost hundreds of millions to billions of US dollars.

Indian-based technology policy expert Prasanto Roy assessed that this case would strengthen India's concerns about strategic independence. He compared it to the lessons from Russia which lost access to SWIFT after the invasion of Ukraine.

"Even if this decision is corrected or reversed, the Anthropic case shows that there is no foreign LLM that is truly geopolitically neutral," Roy said.

LLM or large language model is a large language model that is the basis of AI services such as chatbots. According to Roy, American AI models will always be tied to American geopolitics.