Not All AI Tools Are Created Equal When It Comes to Privacy
Not all AI tools handle your data the same way.
Which version you use, and whether you've changed one setting you may not know exists, determines whether your conversations train future AI models, how long your data is retained, and whether a real human at the company might review what you typed.
This matters enormously for nonprofits. And understanding it doesn't require a technology background. It just requires knowing where to look.
The Big Picture: Two Worlds
The AI privacy landscape has split into two distinct categories. Consumer plans, which are free or low-cost individual subscriptions, generally treat your data as something that can improve their products. Enterprise plans, which involve formal contracts and data processing agreements, treat your privacy as something they are legally obligated to protect.
The problem for most small nonprofits is that they are almost always on consumer plans, often without realizing what that means.
ChatGPT (OpenAI)
On free and most paid consumer plans, OpenAI may use your conversations to train future AI models by default. To stop this, go to Settings, then Data Controls, and disable "Improve the model for everyone."
On ChatGPT Team, Enterprise, and API plans, your data is not used for training by default. OpenAI states explicitly on its business privacy page: "We don't train our models on your organization's data by default." (Source: openai.com/business-data)
The important nuance: ChatGPT Plus users must still manually disable training. Paying for ChatGPT Plus does not automatically protect your data.
Claude (Anthropic)
Anthropic built its early reputation as the privacy-conscious AI company. In September 2025, that changed. Anthropic updated its consumer terms to allow training on conversations from Free, Pro, and Max users unless you actively opt out. If you opted in, your conversations can be retained for up to 5 years. If you opted out, the previous 30-day retention window continues. The update was widely criticized by privacy researchers as a significant retreat from Anthropic's earlier stance.
To check your settings: go to claude.ai, then Settings, then Privacy, and look for the training toggle.
On Claude's commercial API and enterprise plans, data is contractually excluded from training. API log retention was also reduced to 7 days as of September 2025. (Source: anthropic.com/news/updates-to-our-consumer-terms)
Gemini (Google)
On the free consumer Gemini app at gemini.google.com, your conversations may be used to improve Google's models by default. You can disable this by turning off Gemini Apps Activity, though doing so also removes your chat history.
On Gemini for Google Workspace business accounts, Google makes a contractual commitment that your data is not used to train AI models. The official Workspace privacy hub states: "Workspace does not use customer data for training models without customer's prior permission or instruction." (Source: Google Workspace Privacy Hub)
The warning: Gemini Advanced, which is a paid individual subscription, is treated as a consumer account, not a business account. Paying for Gemini Advanced does not give you enterprise privacy protections.
What This Means for Nonprofits in Practice
Most small nonprofits are using free or low-cost consumer plans. That's completely understandable. These tools are powerful, they're accessible, and they work.
But there are three things every nonprofit should do regardless of which tool they use:
First, check your privacy settings today. For Claude, go to Settings and then Privacy. For ChatGPT, go to Settings and then Data Controls. For Gemini, look for Gemini Apps Activity. Opting out of training takes about 30 seconds and costs nothing.
Second, never put sensitive information into any consumer AI tool regardless of your settings. Client names, immigration status, case notes, medical information, and financial details should never go into a consumer AI tool. This rule applies whether your training is on or off, because settings can change, companies update their terms, and the risk simply isn't worth it.
Third, if your organization handles highly sensitive data, the upgrade to an enterprise or business plan is worth evaluating. The privacy protections are contractual rather than just policy-based, which is a meaningful difference.
The Honest Bottom Line
None of the major consumer AI tools are fully private by default. All of them offer stronger protections at the enterprise level. The gap between free consumer use and paid enterprise use is not primarily about features. It is primarily about privacy.
For most small nonprofits, the practical answer is: use the free tools thoughtfully, opt out of training, and never put sensitive information in. That gets you most of the way there without spending a dollar.