Data never has to leave your AWS region. That’s the practical headline for enterprise teams who’ve been evaluating OpenAI Codex and running into data residency objections. As of June 1, OpenAI’s frontier models and Codex are generally available on Amazon Bedrock, confirmed directly from OpenAI’s newsroom. GPT-5.5 and GPT-5.4 are included in the launch.
Per AWS documentation, the integration routes through a `bedrock-mantle` endpoint and offers data residency within the customer’s chosen AWS region, meaning inference requests and outputs stay inside the regional boundary. Per AWS, integrations include VS Code, JetBrains, and Xcode. These specific technical details come from AWS’s documentation, which wasn’t fully accessible for independent review; they carry AWS as the source. Data residency controls are a platform capability claim from AWS, confirm your specific configuration against AWS’s published documentation before treating this as a compliance guarantee.
Codex already had roughly 4 million weekly active users, a figure reported in May and corroborated by earlier TJS coverage. The Bedrock integration doesn’t change what Codex does, it changes where it can be deployed and under whose infrastructure governance it operates.
Analysis
Bedrock data residency is an infrastructure capability claim from AWS, not a compliance certification. Verify that your region's Bedrock deployment holds the specific certifications your regulatory framework requires (FedRAMP, HIPAA, SOC 2) before treating regional data routing as a compliance guarantee.
Who This Affects
Why this matters for regulated industries
Finance, healthcare, and government teams face a consistent objection to third-party AI developer tooling: inference data crossing organizational or jurisdictional boundaries. Running Codex inside Bedrock’s regional data architecture answers that objection directly, at least at the infrastructure level. Whether it satisfies specific compliance frameworks, SOC 2, HIPAA, FedRAMP, sector-specific regimes, depends on the customer’s configuration and AWS’s specific certifications for Bedrock in their region. Don’t assume Bedrock residency equals regulatory compliance; verify the certification scope.
GPT-5.5’s Intelligence Index score of 60.24 is attributed to independent evaluation in the source data. That score isn’t visible in the fetched source content, so it carries qualified attribution here.
The catch is pricing transparency
Pay-per-token pricing through Bedrock hasn’t been confirmed from available sources. Enterprise teams evaluating total cost of ownership need Bedrock-specific pricing before making deployment decisions, OpenAI’s standard API rates may not be the same as Bedrock pass-through pricing. Check AWS’s published pricing for Bedrock model access directly.
What to Watch
What to watch
AWS’s published certification scope for Bedrock (FedRAMP, HIPAA, etc.) determines whether this deployment path is viable for specific regulated-industry use cases. Watch for whether Bedrock-hosted Codex carries the same enterprise agreement terms as direct OpenAI API access, and whether IDE integration depth matches what Codex users currently have on direct API.
TJS synthesis
This is an enterprise distribution story, not a model performance story. GPT-5.5 didn’t get smarter on June 1. What changed is that it became deployable inside AWS’s infrastructure governance stack, and for regulated-industry teams, infrastructure governance is often the harder constraint than model capability. If data residency was your blocker for Codex adoption, that blocker is now addressable. Verify the certification scope against your specific compliance requirements before committing.