Can You Put Donor Data Into ChatGPT, Claude, Gemini or Copilot? A Plan-by-Plan Privacy Table

The short answer: keep donor data out of any personal AI account, on any service. The business and nonprofit plans of the four major assistants don't train on what you type by default, and they come with a data processing agreement. Even on those plans, put in only the data a task needs, and only after your organization has decided in writing that it's allowed.

Below is what each vendor says, plan by plan, from its own pages as of October 24, 2026. This isn't legal advice. If your organization handles health information or has a privacy officer, take this table to them.

The table

	Trains on your content by default?	People may read chats?	How long it's kept	Data processing agreement
ChatGPT Free, Plus, Pro	Yes; you can turn it off. Thumbs up or down sends that whole conversation for training	Not stated	Until you delete; deleted chats removed within 30 days	No
ChatGPT Business	No	Only authorized staff and contractors, for support, abuse and legal reasons	Admins set it	Yes
Claude Free, Pro, Max	Only if you allow it, or the chat is flagged for safety review	Only with your feedback, or for Usage Policy review	If training is allowed, up to 5 years, de-identified	No
Claude Team	No	Only with your feedback, or for Usage Policy review	Kept while you use the product; custom retention is Enterprise only	Yes, in the Commercial Terms
Gemini app, personal account	Yes, while Keep Activity is on	Yes, "a subset of chats"	18 months by default; reviewed chats up to 3 years	No
Gemini app, Google Workspace (including Workspace for Nonprofits)	No, without your permission	No, says Google	Admins set it, up to 36 months	Yes, the Cloud Data Processing Addendum
Microsoft Copilot, consumer	Yes, unless you opt out	Yes, and "an opt-out of human review is not available"	18 months by default	No
Microsoft Copilot Chat with a work account	No	Paid Copilot has opted out of human-review abuse monitoring	Your organization's retention policies apply	Yes

What each vendor says

ChatGPT. OpenAI's data controls FAQ says that on Free, Plus and Pro, "data sharing is enabled by default," and that you can turn off "Improve the model for everyone." It also warns that giving a thumbs up or down means "the entire conversation associated with that feedback may be used to train OpenAI models." For business plans, OpenAI's enterprise privacy page says: "By default, we do not use your business data for training our models." The same page offers a Data Processing Addendum for ChatGPT Business, Enterprise and the API, and says ChatGPT Business completed a SOC 2 Type 2 audit. It names a HIPAA business associate agreement only for the API and ChatGPT for Healthcare, not for Business. Workspace admins on Business can view and export members' conversations, which is worth telling your staff.

Claude. Anthropic's privacy center says that for commercial products, including Claude for Work (Team and Enterprise), "by default, we will not use your inputs or outputs... to train our models," unless you send feedback or opt in. Its Commercial Terms say "Anthropic may not train models on Customer Content from Services" and incorporate a Data Processing Addendum. On consumer plans, chats are used if you allow it or if they're flagged for safety review, and kept up to 5 years in de-identified form if training is allowed. Custom retention and HIPAA support are Enterprise only; Anthropic says Team plans "can't enable HIPAA."

Gemini. For a personal account, Google's Gemini Apps privacy hub says your activity is used to improve its services, "including training generative AI models," that "a subset of chats are reviewed by human reviewers," and that activity is auto-deleted after 18 months. For Workspace, Google's generative AI privacy hub says "Workspace does not use customer data for training models without customer's prior permission or instruction" and "your content is not human reviewed." That hub doesn't name the nonprofit edition, but Google's June 2025 nonprofit update describes "enterprise-grade data protections for all nonprofits using the $0/user Workspace for Nonprofits offering in the Gemini app and NotebookLM."

Microsoft Copilot. Microsoft has renamed its products: Microsoft Learn now says Microsoft 365 Copilot "is now named Microsoft Copilot" and Copilot Chat is "Microsoft Copilot Chat." With a work or school account, Copilot Chat has "enterprise data protection," prompts and responses "aren't used to train the underlying foundation models," and use is covered by Microsoft's Data Protection Addendum. One catch: web search queries are handled by Bing under different terms and aren't covered by that addendum. Microsoft's consumer Copilot FAQ says conversations are used for training unless you opt out, are kept 18 months by default, and that "an opt-out of human review is not available." That FAQ says it applies to the older Copilot app; a new app arrived on August 18, 2026, and we couldn't find its replacement page.

What this means in practice

1. No personal accounts for organization work. Every personal tier above either trains by default, allows human review, or both. Our nonprofit AI policy template puts this in writing.
2. The business tier is affordable. As of October 2026, ChatGPT Business and Claude Team are $8 a user a month for verified nonprofits, and the Gemini app comes free with Google Workspace for Nonprofits. Our AI tools discount comparison has the sources and eligibility, including the open question for churches.
3. Even on a business plan, minimize. Many jobs need no donor data. Drafting an appeal needs the story, not the list. A donor brief before a meeting needs one person's notes, not your whole database.
4. Some data stays out entirely. Prayer requests, counseling notes, health information and anything about minors. On our own systems, the rule is that donor data and passwords never go into notes, articles or prompts that leave our systems.
5. Health data needs more. None of the $8 tiers is set up for HIPAA work. If you handle health information, talk to your privacy counsel before using any AI tool with it.

What we couldn't confirm

Whether OpenAI reviews consumer chats with people isn't stated on the pages we read. Anthropic doesn't say on those pages whether its consumer training setting starts on or off. Google's Workspace privacy hub doesn't name the nonprofit edition directly. And Microsoft's privacy page for its new consumer Copilot app wasn't findable. We'll update this page when we can confirm them.

For what AI agents can run with these safeguards in place, read AI agents for nonprofits. Before adding any AI subscription, the free audit checks what you already pay against nonprofit pricing.