Your AI assistant processes the information needed to answer your prompt. Depending on your provider, account, settings, and connected features, it may also log or retain the conversation, make records available to your employer’s administrators, or send a query to a separate web-search service. “Not used to train models” does not mean “never stored” or “never accessible.”
What happens to business data in an AI assistant?
When you submit a prompt, the service processes it and relevant context—such as an attachment or permitted work file—to generate a response. Other handling is governed by separate rules. Four questions matter:
- Processing: What information does the service use to answer?
- Training: Does the provider use prompts, files, or responses to improve its models?
- Logging and retention: Does it keep interaction records, and for how long?
- Access: Can your employer’s administrators or other authorized people search, audit, or review those records?
These are not interchangeable privacy claims. For example, Microsoft says work or school Copilot Chat interactions are not used to train foundation models, while also explaining that prompts and responses are logged and may be available to IT administrators. Microsoft’s enterprise data protection documentation describes those distinctions alongside other data-handling details.
Does the account or plan change the answer?
Yes. A personal account, an organization’s business account, and an enterprise or education plan can have different terms and controls—even when the product name is similar. Confirm that you are signed into the organization-approved account and that the exact feature you intend to use is covered by its terms.
#1 Best Overall
OpenAI: ChatGPT personal workspaces, business offerings, and API
OpenAI says data from ChatGPT Business, Enterprise, Edu, ChatGPT for Healthcare, ChatGPT for Teachers, and its API platform is not used by default to train or improve its models. That commitment covers inputs and outputs; it is not a promise that content is never stored. OpenAI also describes encryption, retention controls for qualifying organizations, access management, and data-residency options for eligible services. Which controls apply depends on the service and plan. See OpenAI’s business data page.
OpenAI distinguishes those offerings from personal Free, Plus, and Pro workspaces: data sharing for model improvement is on by default in personal workspaces, but can be turned off for new conversations. The setting concerns training, not retention. Details are in OpenAI’s Data Controls FAQ.
Rank #2
For ChatGPT Business, members have separate chat histories and do not automatically see one another’s chats; sharing a chat link is a deliberate sharing mechanism. Workspace spend metrics do not automatically expose members’ full private chat histories. See OpenAI’s ChatGPT Business workspace guidance.
Microsoft: work or school Copilot Chat and Microsoft 365 Copilot
For Copilot Chat signed in with a work or school account, Microsoft says prompts and responses are not used to train foundation models. It also says prompts, Bing queries triggered by prompts, and responses are logged, and IT administrators can use Microsoft search and audit tools to view logged information. A web query follows a separate path: Microsoft describes Bing as an independent controller with different handling terms. Read Microsoft’s Copilot enterprise data protection documentation for the scope and qualifications.
Microsoft 365 Copilot interaction history includes prompts and responses and is stored in line with the organization’s Microsoft 365 contractual commitments. Microsoft says this content is encrypted at rest, is not used to train foundation large language models, and can be searched or governed with Microsoft Purview. See Microsoft’s Microsoft 365 Copilot privacy documentation.
Microsoft says Copilot can inherit identity, permissions, sensitivity labels, retention policies, and audit settings, with exact controls varying by subscription. Its documentation also cautions that agents have their own terms and privacy statements. Organizations with strict geography requirements should check the current scope of the chosen model and account: Microsoft says Anthropic models are currently excluded from the EU Data Boundary when applicable. These details are covered in Microsoft’s enterprise data protection documentation.
Google Workspace with Gemini
Google says qualifying Workspace business and enterprise users receive enterprise-grade protections in the Gemini app: submissions are not used to train models or reviewed by humans, interactions stay within the organization, and existing Workspace protections such as data-region policies and data loss prevention (DLP) apply. Those assurances are specific to qualifying Workspace use. Without a qualifying Workspace edition, consumer terms apply; Google says consumer chats may be reviewed and used for product improvement. Check the organization’s edition and the terms for the exact service. See Google Workspace’s Gemini privacy FAQ.
Workspace administrators can configure Gemini conversation history to retain chats for 3, 18, or 36 months. Conversation history is on by default; if it is turned off, existing chats may remain in user accounts for up to 72 hours for service provision and feedback processing. These are Google Workspace configuration details, not general retention periods for every Gemini product. See Google Workspace Admin Help on Gemini conversation history.
Gemini’s access to Workspace data follows the user’s own access. Administrators may limit access to Gemini or Workspace data, and content owners’ sharing settings still apply. A business plan cannot compensate for files that have been shared too broadly. See Google Workspace’s guidance on Gemini access to Workspace data.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What should you check before using an assistant for work?
- Confirm the account. Make sure you are signed into your organization’s approved business account, not a personal workspace.
- Check the exact plan and feature. Read the applicable service terms, data-processing provisions, and retention policy. A general promise for a provider or product does not necessarily cover every plan, agent, connector, or model.
- Trace the data in both directions. Check how the service treats prompts, attachments, retrieved files, responses, and feedback. Ask whether each is used for training, logged, retained, or handled differently.
- Find out who can access records. Ask whether administrators can view, search, audit, export, or delete interaction history, and how retention is configured.
- Review permissions and connections. Check what the assistant can access through work files, shared drives, connectors, and agents. Tighten excessive sharing and inspect individual agent terms.
- Check separate data paths. Verify the terms and geography for web searches and third-party integrations rather than assuming they follow the same rules as the main assistant.
- Follow your organization’s data policy. Do not submit information that policy, client obligations, or applicable requirements prohibit sending to the selected service.
How to compare business AI data protections
Do not reduce the decision to “private” versus “not private.” Compare the specific product and contract against the questions that matter to your organization:
- Are prompts, attachments, retrieved content, and responses excluded from model training?
- What interaction data is logged or retained, for how long, and under which policy?
- Who can review, search, audit, export, or delete the records?
- Where is data processed or stored, and does the service meet your residency requirements?
- Which identity, permission, sensitivity-label, retention, and DLP controls apply to your subscription?
- Are web search, connected apps, and agents covered by separate terms?
Provider terms, eligible plans, administrator settings, data-region coverage, and service names can change. Verify the current documentation and your organization’s configuration before relying on a particular control.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
