MUMBAI, Oct. 5
Anthropic has introduced in-country inference for its Claude artificial intelligence models in India through Amazon Bedrock, allowing banks, government bodies and other regulated organisations to process AI requests on servers located within the country.
The facility covers Claude Opus 5, Sonnet 5 and Haiku 4.5 and addresses one of the major concerns that has limited wider adoption of generative AI among organisations handling sensitive or regulated data.
Requests submitted to Claude through the India endpoint will now be processed within India, giving organisations greater assurance over data residency.
The move could be particularly important for financial institutions, government agencies and large enterprises that have been reluctant to move sensitive AI workloads into full-scale production when processing takes place outside the country.
Anthropic said India is already one of Claude’s largest global markets and has one of the most technically sophisticated developer communities using its models.
Indian organisations had repeatedly sought the ability to run Claude inference locally because of regulatory and data-handling requirements.
Anthropic India Managing Director Irina Ghose said the company had committed in August to bringing Claude inference to India and had now delivered on that commitment.
The company expects the capability to help customers in regulated sectors move Claude from experimental projects to more critical business applications.
Running inference within India does not merely change where an AI service is accessed. It determines where the processing of prompts and responses occurs, an important consideration when organisations are working with confidential financial, government, corporate or customer information.
For enterprises, that can affect internal risk assessments, compliance requirements and decisions about which workloads can be entrusted to generative AI.
The Claude deployments available through Amazon Bedrock will also provide audit trails and access controls, which Anthropic said are required by risk and compliance teams before AI systems can be deployed in production environments.
AWS India and South Asia President Sandeep Dutta said data residency had become one of the factors determining whether regulated companies could move AI beyond pilot projects.
With the new capability, Claude requests through the India endpoint will be processed on servers based in India.
Banks, insurers and public institutions are among the organisations that could benefit most directly from the change because of their stringent requirements around data governance and regulatory compliance.
Several major Indian companies tested the service before its wider availability.
Reliance and CRED were among the early customers participating in the private preview, according to Anthropic.
Customers testing the system highlighted benefits including regulatory alignment, lower latency, performance and resilience from processing Claude workloads locally.
Tata Consultancy Services also sees growing demand for enterprise AI among companies operating in regulated sectors.
TCS Chief AI and Services Transformation Officer Amit Kapur said Anthropic’s in-country inference capability could help enterprises address evolving requirements around data residency and governance.
The development also comes as Indian companies adjust their data strategies to the country’s evolving privacy regime.
Mahindra Group Chief Technology Officer Mohit Kapoor said the Digital Personal Data Protection Act had made data sovereignty an important business consideration and that keeping data and AI workloads within Indian borders would allow the group to expand its use of Claude with greater confidence.
The National Payments Corporation of India is another organisation incorporating Claude into its artificial intelligence strategy.
NPCI is developing AiNxt, an agentic AI platform using Claude, with the objective of bringing secure and governed artificial intelligence into enterprise-scale operations.
The launch strengthens Anthropic’s growing presence in India.
The company opened its Bengaluru office in February and has since expanded its local team, working with enterprises, technology companies and partners.
India also recorded some of the highest registration numbers globally when the Claude Partner Network was launched in March.
In August, Anthropic held its first Claude Certification for Partners programme in Bengaluru, with certification available for up to 5,000 people.
The addition of domestic inference marks a more significant step because it changes the infrastructure available to organisations deploying Claude rather than simply expanding the company’s commercial presence.
Generative AI adoption within enterprises has often moved faster in experimentation than in production.
Companies can test models using non-sensitive information relatively easily, but deploying them for customer records, financial information, internal corporate data or government workloads introduces substantially more demanding security and compliance requirements.
Data residency addresses one of those barriers.
By allowing Claude requests to be processed within India through Amazon Bedrock, Anthropic and AWS are targeting organisations that need greater certainty over where sensitive workloads are handled.
The development could consequently intensify competition in India’s enterprise AI market, particularly among providers seeking business from banks, insurers, government institutions and large corporations.
For Anthropic, India is increasingly becoming more than a market for Claude users and developers. With a local office, expanding partner network, certification programmes and now in-country inference, the company is building infrastructure aimed at supporting enterprise-scale deployment.
The key test will be whether data residency helps Indian organisations move generative AI from controlled pilots into core business systems.
For sectors where the location and governance of data have been major barriers, Claude’s ability to process requests within India removes at least one of the most significant obstacles.
