Guide OpenAI · Chapter 6

    Data boundaries: where ChatGPT, Codex and the API process.

    The OpenAI API processes per project in the region you choose, with retention options after approval; ChatGPT and Codex follow the workspace, each with date and source.

    As of: October 2026 · Editor: Marcel Haas

    • Europe region
    • Zero Data Retention
    • Compliance API
    • ChatGPT Enterprise
    • Codex Cloud
    • FADP

    May it go there?

    Data class against platform.

    Every cell says whether a data class may be processed on a platform, needs a documented decision, or has no business being there.

    Data classChatGPT BusinessProcessing: Terms of the Business plan, sign-in through Entra ID, no training on contentChatGPT EnterpriseProcessing: Compliance API, SCIM, data residency and retention rules as agreedCodex and Codex CloudProcessing: Connected repository with its permissions, environments and secrets as configuredAPI project, Europe regionProcessing: Listed endpoints in the region, approved retention optionAPI project, globalProcessing: All endpoints, standard retention with logs for up to 30 daysAgents APIProcessing: Not listed for the Europe region
    Public contentWebsite, brochures, published reports
    applies
    applies
    applies
    applies
    applies
    applies
    Internal content without personal referenceDirectives, templates, code without secrets
    applies
    applies
    appliesCode and configuration without secrets in prompts
    applies
    applies
    applies
    Personal dataEmployees, customers, members
    documented decisionOnly with a decision; without residency and the Compliance API
    documented decisionWith recorded residency, retention rules and the Compliance API
    documented decisionTest data without personal reference in the repository; production data only after a decision
    appliesWith an approved retention option, contract and checked endpoints
    no
    noOnly development and test data
    Sensitive personal data and professional secretsHealth, mandates, patient records
    no
    documented decisionDecision of the data protection officer, residency and retention recorded
    no
    documented decisionZero Data Retention, endpoints checked, decision of the data protection officer
    no
    no
    Legend:appliesdocumented decisionno

    As of October 2026, from OpenAI's data controls on developers.openai.com. Our frame for the start, not legal advice; we record the workspace residency with OpenAI before the start, and the decision per project is taken with your data protection officer.

    01

    Every product has its own place.

    You set the data region per API project when creating it, together with keys and spending limits. The Europe region, which OpenAI labels EEA and Switzerland, stores and processes in the region for the listed endpoints and requires a retention option approved by OpenAI; the Agents API is not listed there. The ChatGPT workspace follows the terms of your plan; data residency is among the Enterprise features, and which region a workspace gets we record with OpenAI before the start. Codex Cloud works on the connected repository and respects its permissions; environments, secrets and internet access are defined by administrators.

    02

    What «no training» covers and what logs keep.

    OpenAI has not used API data for training since 1 March 2023 unless you explicitly opt in, and workspace content from Business and Enterprise is not used either. Storage happens all the same: abuse monitoring logs are kept for up to 30 days; Zero Data Retention or Modified Abuse Monitoring take content out of them after approval by OpenAI. In the workspace the retention rules of your plan apply, in Enterprise with periods of your own, and the Compliance API delivers what audit or data protection need to see.

    03

    Administration keeps the evidence.

    In ChatGPT Enterprise the Compliance API delivers conversations and activities to your tools for audit, eDiscovery and loss prevention, SCIM creates accounts and removes them when someone leaves, and roles control who uses Codex, apps or agents. On the platform every application gets its own project with keys that expire, hard spending limits and the data region; so per application you can show where processing happened and how much. On the Microsoft side, Purview's data security posture view also captures a connected ChatGPT Enterprise workspace, and endpoint DLP sees what is pasted into an AI website in the browser.

    04

    The duties under the data protection act stay with you.

    Whichever product processes, you remain responsible. The register of processing activities names the AI application with purpose, data categories and recipients; with OpenAI, the Data Processing Addendum serves as the processor agreement. Professional secrets, as in law firms, practices and hospitals, need their own review before a document reaches a project or the workspace, and for processing with a high risk the act requires a data protection impact assessment. We do not commit to processing in Switzerland for OpenAI's models; we name the region as OpenAI names it.

    05

    The written decision per project.

    Before the first productive call, one page per project states: product and endpoints, data region, retention option, allowed data classes, who approves exceptions and when the statements are checked again. This sheet belongs in the register of processing activities and is signed with your data protection officer. For the workspace the same sheet exists once: plan, residency, retention, approved apps and connectors, and which data classes may be uploaded.

    What you decide.

    Which data class may go into which product?

    Options

    1. 1Public and internal content without personal reference
    2. 2Personal data of employees, customers and members
    3. 3Sensitive personal data and professional secrets

    Our recommendation: Content without personal reference may go into the workspace, to Codex and into any approved project. Personal data belongs in a project with the Europe region and an approved retention option, in the workspace only in Enterprise with recorded residency; the Agents API stays for development and test data. Sensitive personal data and professional secrets we process at OpenAI only after a decision by your data protection officer and with Zero Data Retention; otherwise they stay with Microsoft in a regional deployment or on your own hardware.

    What we advise against: Personal data in a project without region and retention option because the prototype started that way.

    Europe region or global per project?

    Options

    1. 1Europe region for the listed endpoints
    2. 2Global, with every endpoint and feature

    Our recommendation: Europe for everything that touches personal data, checked per endpoint, because the region does not cover every one. Global for development, tests and data without personal reference, and for the Agents API as long as it is not listed in the region.

    What we advise against: One global project for everything because setup is quicker; the region is fixed when the project is created and therefore belongs at the start.

    Standard retention, Modified Abuse Monitoring or Zero Data Retention?

    Options

    1. 1Standard with abuse monitoring logs for up to 30 days
    2. 2Modified Abuse Monitoring after approval by OpenAI
    3. 3Zero Data Retention after approval by OpenAI

    Our recommendation: Standard for data without personal reference. Modified Abuse Monitoring or Zero Data Retention for personal data, after approval by OpenAI; which option you need we settle with your data protection officer, together with the endpoints the application uses.

    What we advise against: Applying for Zero Data Retention without checking which endpoints and features the application can still use afterwards.

    Who signs off the data route of a project?

    Options

    1. 1Your IT alone
    2. 2Business team and IT
    3. 3Business team, IT and the data protection officer

    Our recommendation: All three, on one page per project: product and endpoints, data region, retention option, allowed data classes, who approves exceptions and when the statements are checked again. This sheet belongs in the register of processing activities.

    What we advise against: A decision that covers the whole vendor. The answers differ per endpoint, region and retention option, so per project.

    Status and sources.

    As of: October 2026

    • As of October 2026

      API data: no training, a Europe region per project.

      OpenAI has not used API data for training since 1 March 2023, and abuse monitoring logs are kept for up to 30 days. The Europe region, which OpenAI labels EEA and Switzerland, stores and processes in the region for the listed endpoints and requires a retention option approved by OpenAI. The Agents API is not listed there.

      Source: OpenAI: Data controls in the OpenAI platform
    • As of 12 October 2026

      Zero Data Retention and Modified Abuse Monitoring need an approval.

      Per project you set keys with an expiry date, hard spending limits and the data region when creating it. OpenAI does not use API data for training unless you explicitly opt in. Zero Data Retention or Modified Abuse Monitoring take content out of the abuse monitoring logs after approval by OpenAI.

      Source: OpenAI: Data controls in the OpenAI platform
    • As of 12 October 2026

      ChatGPT Enterprise brings SCIM, role-based access and the Compliance API.

      Business and Enterprise give teams a workspace with sign-in through SAML SSO; Codex is included in both plans. Enterprise adds user provisioning through SCIM, role-based access, the Compliance API for audit and eDiscovery, and the Connector Registry.

      Source: OpenAI: Business plans
    • As of October 2026

      Codex Cloud starts switched off in Enterprise workspaces.

      Administrators enable use through roles, define environments, secrets and internet access, and distribute rules for local runs in a requirements.toml file. Codex Cloud respects the repository permissions of the source system.

      Source: OpenAI: Codex admin rollout guide

    Frequently asked questions.

    What does the Europe region cover?

    Storage and processing for the endpoints OpenAI lists for it, in the region OpenAI labels EEA and Switzerland, with an approved retention option. Endpoints that are not listed, among them the Agents API, process outside; that is why we check the list per application and record the result in the decision.

    Do OpenAI staff see our prompts?

    By default, prompts to the API stay in the abuse monitoring logs for up to 30 days; Zero Data Retention or Modified Abuse Monitoring take them out after approval by OpenAI. For the workspace the business terms of your plan apply; which option you need depends on the data class and is set in the written decision.

    May client or patient data go into a Europe project?

    Only after a decision by your data protection officer, with Zero Data Retention or Modified Abuse Monitoring and with endpoints the region covers; otherwise such data stays in a regional deployment at Microsoft or on your own hardware. Many organisations begin with anonymised cases and open further data only after the decision.

    What does Codex Cloud see of our code?

    The connected repository with the permissions of the source system, the environment with the secrets administrators store, and the internet as far as it is enabled. In Enterprise workspaces Codex Cloud starts switched off; approval runs through roles, and requirements.toml distributes the rules for local runs.

    What does the Compliance API deliver, and who needs it?

    Conversations and activities of the workspace for audit, eDiscovery and loss prevention, to your own tools; it belongs to ChatGPT Enterprise. Whoever must show evidence to auditors or regulators or wants to investigate incidents needs it; a team on the Business plan works without these exports.

    Do we need a data protection impact assessment?

    If the processing can bring a high risk to the personality of the people concerned, the data protection act requires one. With AI applications that is often the case when sensitive personal data is processed at scale or decisions about people are prepared. We supply the technical details for it, data route, region, retention, access and logs; the assessment itself is carried out by your data protection officer.

    Next chapter · Chapter 7

    Which levers move the AI bill at OpenAI.

    How does the meter run at OpenAI, and what moves the bill?