India is reportedly pushing for local hosting of Anthropic’s Claude AI models over cybersecurity and data sovereignty concerns.
Officials fear the company’s Mythos model could rapidly identify vulnerabilities in critical systems such as banking and UPI infrastructure.
The government is also in talks with the US to explore participation in Anthropic’s Project Glasswing security initiative.
The Indian government is pushing for sovereign hosting of Anthropic’s Claude AI models in India amid rising cybersecurity concerns surrounding the company’s Mythos model, according to a MoneyControl report.
Officials argue that advanced AI systems used in sensitive sectors such as banking, telecom and critical infrastructure cannot operate on foreign-hosted infrastructure due to jurisdictional, compliance and national security risks.
1 May 2026
Get the latest issue of Outlook Business
The move aims to protect India’s digital space from evolving cybersecurity threats.
This comes days after Finance Minister Nirmala Sitharaman directed the Indian Banks Association (IBA) led by Challa Srinivasulu Setty to boost its defenses against Mythos. A panel under Setty, who is also the SBI Chairman, was also constituted by the government to assess risks and come up with mitigating measures related to Mythos.
What Is Mythos And Why Is It a Problem?
Mythos is part of Anthropic's Claude family of AI models, built for complex reasoning and coding tasks. What differentiates it is its ability to find cracks and vulnerabilities in existing code with a speed and precision that no human can replicate.
Cybersecurity experts have raised concerns over the AI tool’s ability help attackers break into systems much faster than before. It took the model just seven weeks to identify 2,000 vulnerabilities that human cybersecurity experts had missed for decades.
Anthropic launched the model through Project Glasswing, providing early access to select technology companies such as Microsoft, Cisco, and Amazon Web Services. The project serves another purpose, reinforcing software security infrastructure before vulnerabilities can be weaponised. It should be noted that no Indian company is part of Project Glasswing.
The Indian Government is engaged in bilateral talks with the US to develop mechanisms that would allow India to participate in Project Glasswing, the Moneycontrol report said.
Stakes for India
The Indian financial system represents a concentrated and consequential risk that would affect hundreds of millions of people. India’s digital payments infrastructure under UPI can easily be scanned and probed for vulnerabilities by Mythos class AI models.
The Indian banking system is also uniquely exposed to risks posed by the model. For Indian banking systems uptime is paramount meaning banking systems cannot be patched quickly.
These one-of-a-kind risks heightens the importance of sovereign data hosting in India.
In May 2025, OpenAI, introduced data residency in multiple Asian markets including India to meet regulatory compliances. This allows customer data to be stored in India but Inference-layer processing still runs on overseas servers.
Anthropic has not yet given any assurance on sovereign hosting arrangements in India, the Moneycontrol report stated.
