What happened: OpenAI announced on November 12 that eligible businesses and developers in India can now run GPT-6 Luna API requests entirely inside India — the processing itself, not just data storage. Why it matters: for Indian banks, insurers and other regulated firms that could not send data abroad for processing, this removes the blocker that kept them off the API.
November 12, 2026
What changed: residency now covers inference
OpenAI already offered data residency in India, but that only governed where customer content was stored. With API inference residency, both halves stay in-country: for supported workloads, the API request is processed in India and the model computation that generates the response also happens in India, with supported customer content stored there too. The initial scope is narrow and specific: direct API usage of the GPT-6 Luna model, through the Responses and Chat Completions APIs.
The conditions attached
This is not a switch every Indian account can flip. Coverage reports based on OpenAI’s documentation say regional processing requires an organization to have active Zero Data Retention, which itself is subject to OpenAI approval — a weaker abuse-monitoring setting is not enough. The India region supports Standard pay-as-you-go processing only, OpenAI-hosted tools are not available under regional processing, and customers start by creating a new API project with India selected for both storage and processing, then routing requests through the India regional endpoint. One report puts a 10% premium on GPT-6 Luna regional processing; buyers should confirm current pricing in OpenAI’s docs before budgeting.
Part of a wider residency push
India is not a one-off. OpenAI describes India, Australia and Japan as the most recent additions to its API inference residency offering, which tells you where enterprise demand for in-country processing is loudest. The pitch, as OpenAI’s head of enterprise for India put it in launch coverage, is aimed at businesses that have held back from deploying AI on important work because the information involved had to be processed within the country — document summarization, response drafting and information analysis are the example workloads.
Why it matters
Data-residency checkboxes are quietly becoming the real enterprise AI battleground. Storage residency was the easy half; inference residency is what regulated industries actually need, because the prompt itself is often the sensitive part. The trade-off is real — one model at launch, no hosted tools, an approval gate, and possibly a price premium — but for an Indian financial-services firm, a slightly limited model that legal can approve beats a frontier model it cannot touch. Expect every major lab to keep expanding in-country inference, region by region.
Frequently asked questions
Which model does OpenAI’s India inference residency cover?
At launch, only GPT-6 Luna, through the Responses and Chat Completions APIs, for direct API usage.
What do customers need to qualify?
An organization needs active Zero Data Retention (subject to OpenAI approval), a new API project set to India for both storage and processing, and Standard pay-as-you-go billing.
Which other countries have OpenAI inference residency?
India joins Australia and Japan among the most recent locations added to the offering.
Sources: The Hindu BusinessLine, DQ India, FoneArena