ChatGPT Enterprise is private enough for most marketing work, but not for every use of customer data. OpenAI does not train on Enterprise data by default, encrypts it, signs a Data Processing Addendum, lets admins set retention and, for new customers, stores content in a chosen region. The conversations still live on OpenAI's systems, OpenAI runs them through automated safety classifiers, and authorized OpenAI staff can access them for set reasons. For support replies on order history or segmentation on customer profiles at volume, a brand needs more than the Enterprise plan.
What does ChatGPT do with company data on each plan?
OpenAI's terms differ by product, and the gap between a personal account and a company workspace is larger than the gap between Business and Enterprise. The table sets out the products a marketing team is likely to use.
| Question | ChatGPT Free, Plus, Pro | ChatGPT Business | ChatGPT Enterprise | OpenAI API |
|---|---|---|---|---|
| Used to train OpenAI's models? | Yes by default, with an opt-out; feedback can bring a conversation into training | No by default, unless the customer opts in | No by default, unless the customer opts in | No since 1 March 2023, unless the customer opts in |
| Who at OpenAI can access conversations | Under consumer terms | Authorized staff for engineering support, abuse investigation and legal compliance; contractors for abuse review | Authorized staff only to resolve incidents, recover conversations with permission, or where the law requires | Under API data controls |
| Who sets retention | The user, by deleting chats | Workspace admins. Deleted or unsaved conversations are removed within 30 days | Workspace admins, with custom retention policies. Deleted conversations are removed within 30 days | OpenAI keeps inputs and outputs up to 30 days. Zero data retention on request for eligible endpoints |
| Data stored in a chosen region | No | No | Yes, for new customers, in ten regions at no additional cost | Yes, for eligible customers |
| SOC 2 Type 2 | No (Plus, per OpenAI's pricing page) | Completed | Completed | Audited and certified |
| Data Processing Addendum for GDPR | Not listed by OpenAI | Available | Available | Available |
Across plans, OpenAI encrypts data at rest with AES-256 and in transit with TLS 1.2 or higher. On the business side, the training default is stated plainly: by default, OpenAI does not use inputs or outputs from ChatGPT Business, ChatGPT Enterprise or the API to improve its models, and on those plans "response feedback (submitting thumbs up or thumbs down) does not permit training on the associated conversation." ChatGPT Business vs Enterprise compares the two plans on price, admin controls and residency, and private AI vs public AI compares them with a model running in the brand's own account.
What does ChatGPT Enterprise not cover?
Enterprise controls what OpenAI may do with the workspace's data. It does not keep that data inside the brand's own systems. OpenAI's enterprise privacy page limits OpenAI's access to Enterprise conversations to three cases:
Authorized OpenAI employees will only ever access your conversations for the purposes of resolving incidents, recovering end user conversations with your explicit permission, or where required by applicable law.
The same page says OpenAI may run any business data through automated content classifiers and safety tools. For most marketing work that is acceptable. For a support team pasting customer names, addresses and order histories into prompts every day, it means those records sit on OpenAI's systems, under OpenAI's access rules, until the 30-day deletion window runs out. On ChatGPT Business the access rules are wider: OpenAI's staff can reach conversations for engineering support and abuse investigation, and specialized contractors can review them for abuse and misuse.
Data residency narrows where content is stored, but not everything stays in region. OpenAI's residency article says account data, billing details and other metadata are out of scope, and that processing such as authentication and routing may still happen outside the chosen region.
Three gaps sit outside the plan altogether:
- Personal accounts. A staff member who uses a personal ChatGPT account is on consumer terms, where content trains OpenAI's models unless the user turns off "Improve the model for everyone", and where submitting feedback can bring the whole conversation into training even after opting out. The Enterprise admin console cannot see or control it.
- Admin access inside the brand. On Business, OpenAI notes that workspace admins can view, access, export and delete end user conversations. On Enterprise, admins can access an audit log of conversations through the Enterprise Compliance API. Either way, customer data pasted by one team is visible to whoever administers the workspace.
- Connected apps. OpenAI says it does not train on data accessed from apps by default, but a connected CRM or drive still sends records into OpenAI's service when the assistant reads them.
The risks of using public AI tools with customer data covers these gaps across vendors, with the steps to close them in order.
When is ChatGPT Enterprise enough, and when is it not?
It is enough when prompts carry no customer records or only small amounts: product copy, ad copy, research, internal documents and analysis of aggregated sales data, with the Data Processing Addendum signed and personal accounts banned for company work. It is not enough when a workflow sends customer records through the model at volume, such as support replies drawing on order history or segmentation on customer profiles, or when the data has to stay in the brand's own cloud account.
There are two ways to close that gap. One keeps Enterprise and removes personal data before it reaches the model, with access control and audit logging around it, which is what AI data security on AWS covers. The other moves customer-record work to a private LLM in the brand's own account, where no outside model vendor stores the prompts. Brand Media Works, an AI consulting and marketing data engineering firm for ecommerce brands, builds both. Self-hosted LLM vs ChatGPT Enterprise compares the two options on cost and effort, and private AI for business covers when a marketing team needs one.
Frequently asked questions
Does ChatGPT train on my business data?
It depends on the plan. On ChatGPT Free, Plus and Pro, content is used to train OpenAI's models unless the user opts out. On ChatGPT Business and Enterprise, it is not used for training by default, and thumbs-up or thumbs-down feedback does not permit training on the conversation. Data sent to the OpenAI API has not been used for training since 1 March 2023 unless the customer opts in.
Can OpenAI staff read our ChatGPT Enterprise conversations?
Only in limited cases. OpenAI says authorized employees access Enterprise conversations only to resolve incidents, to recover conversations with the customer's explicit permission, or where the law requires it. On ChatGPT Business, staff can also access conversations for engineering support and abuse investigation, and contractors bound by confidentiality may review them for abuse and misuse.
Can ChatGPT Enterprise keep our data in Europe?
Yes, for new Enterprise customers. OpenAI offers data residency in ten regions, including Europe, the UK, the US and the UAE, at no additional cost on Enterprise. ChatGPT Business does not offer it.
Does OpenAI sign a Data Processing Addendum for ChatGPT Enterprise?
Yes. OpenAI says it can execute a Data Processing Addendum for ChatGPT Business, ChatGPT Enterprise and the API, in support of GDPR and other privacy laws.