OpenAI now processes AI requests inside India
AI IndiaNovember 13, 20264 min readBy Digitalaikart Editorial

OpenAI now processes AI requests inside India

#OpenAI#IndiaAI#DataLocalisation#EnterpriseAI

TL;DR

01.What OpenAI has switched on

OpenAI said this week that its developer platform can now handle supported work entirely inside India. Until now, an Indian customer could ask for data to be stored in the country, but the actual processing of a request still happened on servers abroad. From now, both the API request and the model computation that writes the answer take place in India, and supported customer content is stored here as well. The company calls the option API inference residency, and India is one of its newest markets.

At launch the option covers one model, GPT-6 Luna, and only direct use of the API through the Responses and Chat Completions interfaces. Those accept text and image inputs and return text. That is enough for ordinary office work such as summarising long documents, reading scanned files, drafting replies for staff to review, and answering questions over internal material.

02.Storage and processing are different things

Keeping a file in India does not decide where a model reads it. A firm could store a contract on an Indian server and still have it processed abroad. Inference residency closes that gap. Nitin Bawankule, who heads enterprise business for India at OpenAI, said that some Indian businesses have not fully deployed AI for important work because they need the information involved to be processed within the country. He said the change gives more businesses a way to put these tools to work and gives Indian developers and startups more scope to build.

For financial services the question is practical rather than technical. Several Indian sectors work under rules that decide whether customer information may be handled outside the country, and a model call that leaves India can be enough to stop an application from going live. OpenAI said the rollout is aimed at organisations with specific localisation requirements, and financial services customers are the clearest example it gave.

03.Who can use it, and what it costs

Access is not automatic. An organisation needs an approved Zero Data Retention arrangement, which OpenAI grants case by case, and modified abuse monitoring on its own is not enough. Eligible customers then create a new API project, choose India for both storage and processing, and send requests to the India regional endpoint. Only standard pay-as-you-go processing is offered, and OpenAI lists a ten per cent premium for GPT-6 Luna regional processing, so the usual Batch and Flex discounts do not apply in India for now.

That premium matters for anyone planning volume. A team that processes thousands of invoices or scanned forms a month should work out its cost at the regional rate before committing, rather than assuming the rates it already pays. It also means the option is best judged workload by workload: high-value documents that must stay in the country, not every small task inside an application.

04.What it does not cover

The change is narrower than it first sounds. It does not apply to ChatGPT or to other OpenAI models, and OpenAI-hosted tools cannot be used for regional processing in India. System data such as billing details, usage statistics and some metadata may still be processed or stored outside the selected region, even when customer content stays here. Teams moving an existing product need to check whether their whole workflow fits the regional setup.

A document review pipeline may fit if it sends supported inputs and expects text back. One that depends on hosted tools, or that passes files to another vendor before or after the model call, will need redesign or must stay outside the regional deployment. Local inference answers where the model runs. It does not answer where your own application keeps its logs and outputs.

05.What this means for Indian readers

For students and early-career professionals the signal is about skills. Banks, insurers and hospital groups that could not use foreign AI earlier now have a route to production work, and that work needs people who can build retrieval, testing and audit trails around a model, not only people who can write a good prompt. Junior roles in Indian AI teams are likely to grow around the compliance work as much as around the model itself.

For small businesses the near-term answer is patience. The eligibility bar and the premium mean this is not a change for a shop or a small agency today. The direction is what matters: the option arrived as part of a wider India programme that includes a partnership with the Tata Group for domestic data centres, and India joins Australia and Japan in the latest expansion. More models and cheaper processing are the reasonable expectation from here.

Frequently asked questions

Is my ChatGPT data now processed in India?
No. The local option applies to the API only, and at launch it covers a single model, GPT-6 Luna. ChatGPT plans and other OpenAI models are not part of this rollout.

Do I need special approval to use it?
Yes. Your organisation needs an approved Zero Data Retention arrangement, and requests must go to the India regional endpoint from a project set to India for storage and processing.

Will it cost more than the usual API rates?
OpenAI lists a ten per cent premium for GPT-6 Luna regional processing, and only standard pay-as-you-go is offered, so Batch and Flex discounts are not available for these requests.

Found this useful? Share it

Source