
OpenAI starts processing eligible API workloads inside India
Indian companies gain local AI processing for selected sensitive workloads.
Overview
Indian businesses can now run some requests in India instead of sending supported requests and model work abroad. The launch covers through the and . Supported content is stored in India, and answers are produced on computer systems there. This could help regulated companies follow local data rules and use AI for sensitive work with more confidence.
The option does not keep an entire AI operation in India. Customers must create a new API project, choose India for storage and processing, use , and meet OpenAI’s access rules. These include supported workloads and approved payment or plans. Storage can be local, but model processing, other models, system data, CPU work, and connected services may still be handled elsewhere. Companies must check each workflow rather than assume all data stays in India.
Timeline
OpenAI clarifies that inference residency does not localize every AI workflow
specifies where supported information is stored, while inference residency specifies where GPU computation on in-scope content occurs. The India commitment does not guarantee that all , system data, every model, or every part of an organization’s wider AI workflow remains in India.
OpenAI restricts India residency to eligible customers and supported workloads
The capability is available only to eligible customers using supported workloads and direct API access. OpenAI’s controls identify processing and, for the relevant regional setup, as conditions tied to access.
OpenAI requires customers to create India-configured API projects
Customers must create a new , select India for both data storage and processing, and send requests through the . These settings determine whether supported requests are routed for local handling.
OpenAI enables India-based GPT-6 Luna inference through two direct APIs
Direct API customers can use the India capability through the Responses API and . The support therefore applies to applications making qualifying direct API calls, rather than automatically covering every OpenAI interface or product.
OpenAI limits India inference residency to GPT-6 Luna input and output
The launch initially supports for both input and output. This makes the rollout narrower than a blanket India-local option for every OpenAI model or workload.
OpenAI launches local API processing and inference in India
OpenAI announced that eligible businesses and developers can run supported within India. For those workloads, supported customer content is stored locally, API requests are processed locally, and the model computation that generates responses occurs in India.