Regional AI
AWS Keeps Two OpenAI Models Inside Its India Routing Boundary
AWS said on August 27 that Amazon Bedrock's India profiles for OpenAI GPT-5.6 Terra and Luna route inference only between Mumbai and Hyderabad. The boundary covers processing and routing, while flagged content can be retained for offline abuse detection and legal compliance remains workload-specific.
Citation-ready: Amazon Bedrock's India geographic profiles for OpenAI GPT-5.6 Terra and Luna route inference only between the Mumbai and Hyderabad AWS Regions.
What happened and why it matters
No. They keep inference routing within India across Mumbai and Hyderabad, not within one city, and AWS discloses a flagged-content retention exception plus separate global profiles that can route worldwide.
Official AWS regional-inference documentation
Primary reference: AWS OpenAI models for India geographic inference. Kaleido Field checked the event date and the article's attributed facts against this source.
| Source date | August 27, 2026 |
|---|---|
| Checked by Kaleido Field | August 30, 2026, 08:18 CST |
| Source function | current regional-inference analysis separating model availability, geographic routing, logging location, zero-retention default, abuse-detection retention, global profiles, and legal compliance |
The profile is the routing control
The `in.` profiles draw capacity from ap-south-1 and ap-south-2 while keeping processing in India. AWS distinguishes them from `global.` profiles, which can route from an India endpoint to supported commercial Regions elsewhere.
A deployment record should preserve the exact profile ID, source Region, destination Region from CloudTrail, API endpoint, model version, and any fallback path.
Residency is not a blanket compliance verdict
AWS says prompts and outputs are not stored by default, then names an exception for GPT-5.6 content flagged by automated abuse classifiers. It also notes that monitoring stays in the source Region while prompts and results can move between the two India Regions.
Organizations still need to map logs, caches, prompts, outputs, connectors, support access, retention policy, encryption, incident response, and sector rules across the full application.
Evidence boundary
Official product facts: Terra and Luna availability, one-million-token context, text and image input, India-only routing, profile IDs, supported APIs, prompt caching, log locations, CloudTrail inference-region field, and IAM requirements. AWS security statement: zero data retention by default, with content flagged by automated abuse detection retained for offline review. Not established: legal compliance for a specific workload, model quality, latency distribution, cost advantage, or residency of systems outside the Bedrock inference path.
FAQ
What is the practical answer?
AWS said on August 27 that Amazon Bedrock's India profiles for OpenAI GPT-5.6 Terra and Luna route inference only between Mumbai and Hyderabad. The boundary covers processing and routing, while flagged content can be retained for offline abuse detection and legal compliance remains workload-specific.
What source does this article use?
The primary source is AWS OpenAI models for India geographic inference. Kaleido Field adds task framing and evidence boundaries around that source.
Where should the user verify the answer?
Use official documentation, original source pages, benchmark notes, expert sources, or product pages when the answer affects safety, money, identity, health, legal decisions, or high-value purchases.