Organisations that need to run OpenAI models inside a US government cloud boundary now have a supported path: the ChatGPT Amazon Bedrock GovCloud configuration brings the ChatGPT desktop app to Amazon Bedrock in AWS GovCloud (US), giving access to OpenAI frontier models through two components. It is a controlled deployment, not a consumer sign-in — model inference runs through AWS, under AWS identities and network controls, with the guardrails set by administrators before anyone launches the app.
This article explains what is now available, why the GovCloud setup differs from ordinary ChatGPT use, how requests are routed, what it costs, and what the official walkthrough leaves unanswered.
ChatGPT on Bedrock in GovCloud: what is now available
The ChatGPT desktop app is now available on Amazon Bedrock in AWS GovCloud (US). It provides access to OpenAI frontier models through two components. Work completes complex, multi-step tasks and delivers review-ready results, and Codex is a coding agent for developing, testing, and reviewing software.
Work completes complex, multi-step tasks and delivers review-ready results. Codex is for technical work such as editing code, running tests, and reviewing pull requests. The distinction matters for deployment: the two components serve different jobs.
Codex is the OpenAI coding agent for AI-powered software development.
Why GovCloud changes the setup
This walkthrough is intended for organisations whose security architecture requires model inference to use AWS GovCloud (US) identities, endpoints, and regional controls. That single sentence is the whole reason the setup exists: in these organisations, sending prompts to a public OpenAI endpoint is not an option, even if the model itself is approved.
The deployment described in the AWS walkthrough is a controlled local deployment of the ChatGPT desktop app that sends model inference to Amazon Bedrock in AWS GovCloud (US-West). The configuration uses API-mode sign-in, a managed device policy, a named AWS profile, and network restrictions. Those restrictions allow access only to the approved Amazon Bedrock endpoint while disabling other external connections and network-enabled capabilities. The result is a deployment that an agency security team can test and document before approving broader use.
There is an important caveat that should not be buried. Using AWS GovCloud (US) and following the guide do not, by themselves, authorise a workload. The organisation remains responsible for reviewing the selected model, AWS services, Region, endpoint, workstation controls, repositories, credentials, and data flows against its system security plan and authorisation process. GovCloud is a set of controls, not a stamp of approval.
How the ChatGPT Amazon Bedrock GovCloud routing works
The walkthrough configures Codex to route model requests through Amazon Bedrock Mantle in us-gov-west-1, where access can be governed using AWS Identity and Access Management (IAM) and existing AWS GovCloud (US) network controls. The ChatGPT desktop app runs on an organisation-managed computer through the desktop app or CLI, and it uses the configured AWS credentials to securely send requests to the Amazon Bedrock Mantle endpoint in us-gov-west-1.
The main security controls are the computer’s policies and the AWS identity and permissions being used.
Model inference is routed through the Responses API on Amazon Bedrock. The configuration points at the Mantle endpoint for the GovCloud West Region, and the walkthrough’s network policy allows exactly that endpoint while denying OpenAI’s public domains.
Ways to use it: desktop app, CLI, and IDEs
There are three ways to use Codex in this setup. You can use the ChatGPT desktop app, the Codex CLI, and IDE integrations with Visual Studio Code, JetBrains, and Xcode. The desktop app covers Work and Codex; the CLI and IDE integrations cover the coding workflows where developers already work.
For developers, the practical implication is that the tool meets them where they are. Editing code, running tests, and reviewing pull requests can happen inside the IDE, while larger multi-step tasks can be handed to Work in the desktop app. The documented routing through the Bedrock endpoint and IAM-governed identity applies to Codex; whether the walkthrough’s configuration covers Work remains unconfirmed.
Verification is deliberately visible. After launch, the app must show amazon-bedrock as the model provider, and the Codex CLI /status tab confirms the same. If the app prompts for an OpenAI account instead, check that the configuration files are in place before relaunching — the files must precede the first app launch.
Data residency, billing, and zero data retention
For customers with data residency requirements, processing stays within the Amazon Bedrock AWS Region that you select. In this walkthrough that is us-gov-west-1, and the conclusion of the AWS post states plainly that inference stays in the AWS GovCloud (US) Regions.
Billing is simple and worth stating exactly: you pay per token with no seat licenses and no per-developer commitments. Cost therefore tracks usage directly.
Zero data retention (ZDR) is also supported. In this configurable mode, prompts and responses are processed without being written to durable storage by AWS or shared with model providers. The AWS post frames this as helping organisations protect sensitive information and meet stringent data-governance, privacy, and security requirements. Note the scope: it covers durable storage by AWS and sharing with model providers — it is not a blanket claim about all logging.
Configuration walkthrough at a glance
The configuration has three main steps: set up authentication, set up the ChatGPT desktop app configuration files, and then launch, initialise, and verify the deployment. Before starting, make sure the installed version of the ChatGPT desktop app supports Amazon Bedrock in AWS GovCloud (US), and that you have an approved AWS GovCloud (US) account, an IAM identity and approved credentials, access to the approved model, AWS Region, and endpoint, and permission to use them.
Step one is authentication. Select the approved AWS GovCloud (US) account and IAM identity or role with access to the approved model in us-gov-west-1, and obtain credentials through the organisation’s approved mechanism. The walkthrough uses the named profile codex-il5.
Two authentication pathways are supported: an Amazon Bedrock API key, or the AWS SDK credential chain. AWS SSO through AWS IAM Identity Center is supported within the credential-chain pathway. The app first checks for AWS_BEARER_TOKEN_BEDROCK in its environment, and when present, the bearer token takes precedence over profile-based authentication. If no bearer token is available, the app falls back to the standard AWS SDK credential chain.
Step two is the configuration files. You must create two files, config.toml and requirements.toml, before the application is launched. requirements.toml is the administrator-controlled rules file: it sets security limits that users cannot override, for example restricting which permission levels, tools, or settings the app is allowed to use. The baseline in the walkthrough restricts sign-in to API mode, disables browser and computer use, in-app browser, image generation, memories, plugins, and external network access, and allows only the approved Bedrock endpoint.
Then config.toml tells the app how to operate, including which model provider to use (amazon-bedrock), which model to select, and settings such as permissions. The researched values the configuration requires are the provider amazon-bedrock, the Mantle endpoint https://bedrock-mantle.us-gov-west-1.api.aws/openai/v1, the Responses API, region us-gov-west-1, and the named profile codex-il5; the model itself must be an approved model ID, which the walkthrough leaves as a placeholder.
Step three is launch, initialise, and verify. Make sure requirements.toml and config.toml are in place before the first app launch, then start the app and confirm it shows amazon-bedrock as the model provider.
Limits and open questions
Several things the announcement does not settle. The walkthrough uses a REPLACE_WITH_APPROVED_MODEL_ID placeholder, and the supplied evidence does not identify the available model IDs. Organisations must separately approve the model they use, but that approval does not determine which models the service makes available. The walkthrough configures Codex and names us-gov-west-1; whether other GovCloud Regions are supported, and whether Work is covered by the same configuration steps, is not spelled out. The required desktop app version is also a prerequisite to check rather than a stated version number.
Troubleshooting guidance in the post reinforces the limits: if the AWS GovCloud (US) endpoint is reported as unsupported, confirm the exact app release, endpoint, and model with OpenAI.
For teams evaluating this, the honest summary is: the ChatGPT desktop app on Amazon Bedrock in AWS GovCloud (US) gives a controlled way to use OpenAI frontier models and coding assistance, with inference staying in GovCloud Regions, governed by IAM, billed per token, and with zero data retention available. The deployment work — identities, profiles, configuration files, network policy, and verification — is where the security actually lives, and the walkthrough gives security teams a deployment they can test and document before approving broader use.
