Why Seniors Should Think Twice Before Typing Personal Information Into AI Chatbots
AI chatbots like ChatGPT, Google Gemini, Microsoft Copilot, and Claude have quickly become part of everyday life. Many Arizona seniors use them to look up medication questions, draft letters, sort out Medicare options, or simply have someone to “talk” to. They can be genuinely useful. But there is one thing every user should understand before typing a single word: in most cases, what you type does not stay between you and the computer.
How Chatbots Learn
AI chatbots are built by feeding enormous amounts of text into a computer system until it learns to predict and produce language. That training does not stop once the product launches. The companies behind these tools continue improving them, and one of their most valuable sources of new material is the conversations people have with the chatbot every day.
This is not a fringe concern. Researchers at Stanford’s Institute for Human-Centered AI reviewed the privacy policies of six major US AI developers, including Amazon, Anthropic, Google, Meta, Microsoft, and OpenAI, and found that leading AI companies are pulling user conversations for training. As of mid-2026, ChatGPT, Gemini, and Claude all train on consumer chats by default, and a paid plan does not opt you out.
In plain terms, unless you go into the settings and change it, your conversation may become part of the material used to build the next version of the product.
Real People May Read Your Conversations
Many people assume that a chatbot conversation is handled only by machines. That is not always true. Google’s own privacy notice for Gemini states that human reviewers (including third parties) read, annotate, and process your Gemini Apps conversations. Google says it takes steps to protect your privacy as part of this process, for example by disconnecting your conversations from your Google Account before reviewers see them.
That safeguard helps, but it does not erase what you wrote. If you typed your full name, address, or a description of your medical condition into the chat, a stranger may still be reading those words. Reviewed conversations can also stick around: conversations that have been reviewed or annotated by human reviewers are not deleted when you delete your Gemini Apps activity.
Turning Off Training Does Not Undo the Past
Most major chatbots now offer a setting to stop your chats from being used for training. That is worth doing, and step-by-step directions for each major chatbot appear later in this article. But the setting has an important limit. Opt-out settings typically prevent future conversations from entering training, but they do not necessarily retract data already included in a completed training run. Opting out also does not mean your information disappears. Opting out only stops the training. Your data still sits on their servers.
Once information has been typed and sent, you should assume you can no longer fully take it back.
Why This Matters More for Seniors
Older adults are already the most heavily targeted group for fraud in Arizona. As Arizona Attorney General Kris Mayes has put it, fraudsters and scammers take advantage of Arizonans, especially seniors. Any piece of personal information that leaves your control adds to the risk, whether through a data breach at a tech company, a reviewer who should not have access, or details that end up linked together over time.
Health information deserves special caution. Many people assume their medical details are protected the way they are at a doctor’s office. They often are not. Most chatbots aren’t bound by health-privacy laws like HIPAA, and your symptoms can be stored or used to infer things about you.
There is also a newer danger. Scammers now use AI tools themselves to write convincing emails, texts, and phone scripts. The more personal detail that exists about you in digital form, the easier it becomes for a criminal to craft a message that sounds like it came from your bank, your grandchild, or Medicare.
What You Should Never Type Into a Chatbot
Treat every chatbot conversation as if it could be read by a stranger. With that in mind, never enter:
Your Social Security number or Medicare number
Bank, credit card, or account numbers
Passwords, PINs, or security question answers
Your full name together with your home address or date of birth
Specific medical diagnoses, prescriptions, or test results tied to your name
Details about family members, such as grandchildren’s names, schools, or phone numbers
Photos of documents like driver’s licenses, insurance cards, or bank statements
Information about your finances, savings, or estate plans
How to Use Chatbots More Safely
You do not have to give up AI tools altogether. You just need to use them carefully.
Ask general questions instead of personal ones. Rather than typing “I’m 74, I live at 123 Main Street in Sierra Vista, and I take metformin and lisinopril, what are the side effects,” ask “What are the common side effects of metformin and lisinopril taken together?” You get the same useful answer without handing over who you are and where you live.
Turn off training in each chatbot you use. The directions for each major chatbot are in the next section. If you cannot find the setting, ask a trusted family member or a local senior center to help you check.
Use temporary or private chat modes for anything sensitive. Several chatbots offer a mode that does not save the conversation or use it for training. Even in these modes, the rule about not typing personal information still applies.
Remember that a chatbot is not your doctor, lawyer, or banker. Use it to prepare questions or understand general topics, then take your actual personal details to a professional who is bound by confidentiality rules.
Be suspicious of any “chatbot” that asks for personal details. Legitimate AI tools do not need your Social Security number or bank login to answer a question. If a website, pop-up, or app that looks like a chatbot starts asking for that kind of information, close it.
Step-by-Step Privacy Settings for the Major Chatbots
An important note before you start: no setting removes every risk. These settings keep your future conversations out of AI training, but the company still stores your chats for a period of time, and anything already used in training cannot be pulled back. Treat these settings as a second layer of protection. The first layer is still keeping personal information out of the chat box. Menu names change from time to time, so if a step does not match exactly, look for a “Privacy” or “Data” section in the settings.
ChatGPT (OpenAI)
Log in to ChatGPT, click your profile icon, go to Settings, select Data Controls, and turn off “Improve the model for everyone.” The steps are the same in the phone app. For any one-time conversation you would rather not have saved, start a Temporary Chat. Temporary Chats are not used for training and are deleted within about thirty days.
Google Gemini
On a computer, sign in at gemini.google.com, click Settings & help at the bottom left, then Activity. Near the top, click On, then choose Turn off or Turn off and delete activity. On a phone, open the Gemini app, tap Menu, then your profile picture or initial, then Gemini Apps Activity, and tap On, then Turn off or Turn off and delete activity. We recommend choosing “Turn off and delete activity” to clear out older conversations too.
Be aware of two limits. Even with this setting off, your conversations are kept with your account for up to 72 hours so Google can run the service. And chats already reviewed by human reviewers are kept, disconnected from your account, for up to three years. The setting has been on by default for adults, so if you have used Gemini before, assume past conversations may already have been reviewed.
Microsoft Copilot
Copilot has two separate switches, one for typed conversations and one for spoken ones. Turn off both. On copilot.com, select your profile icon, select your profile name, then select Privacy and turn off Model training on text and Model training on voice. In the Copilot app for Windows or Mac, select your profile icon, then Settings, then Privacy. In the phone app, open the menu, select your profile icon, then Account, then Privacy. Depending on your version, the switches may be labeled “Training on conversation activity” and “Training on voice conversations.” Turn off both. If you use Copilot on more than one device, check the setting on each one.
Claude (Anthropic)
Open Claude, click your name or profile icon, go to Settings, then Privacy, and turn off the “Help Improve Claude” setting. This works the same on the website and the phone app. For sensitive one-time questions, Claude also offers Incognito chats, which are never used for training.
Grok (on X, formerly Twitter)
In your X settings, select “Privacy & Safety,” scroll to “Data sharing and personalization,” select “Grok & Third-party Collaborators,” and uncheck the box that allows your public data and your Grok interactions to be used for training and fine-tuning. The same screen also lets you delete your Grok conversation history, and deleted conversations are removed from X’s systems within 30 days except where kept for security or legal reasons.
One catch: X says that pressing the thumbs up or thumbs down feedback buttons lets it use that conversation for training even after you opt out. If you have opted out, skip the feedback buttons.
Perplexity
Perplexity trains on conversations when you are signed in unless you turn off the AI Data Retention toggle under Settings, Account.
Meta AI (Facebook, Instagram, Messenger, WhatsApp)
Meta AI is the hardest to control, and seniors should be especially cautious with it because it appears inside apps many people use every day. As of 2026, there is no comprehensive opt-out for US residents. In addition, starting December 16, 2025, your conversations with Meta AI across Facebook, Instagram, Messenger, and WhatsApp can be used to personalize the ads and content you see, and US users cannot opt out. US users can submit an objection request through Meta’s Privacy Center, but Meta does not guarantee it will be honored.
The safest approach with Meta AI is simply not to use it for anything personal. Your ordinary private messages to friends and family are a different matter, since Meta excludes private messages between friends and family from its generative AI training, but be careful in group chats, where anyone can tag @MetaAI and bring the AI into the conversation.
If You Think Your Information Has Been Misused
If you shared personal information and are worried it has been misused, or if you believe you have been targeted by a scam, contact the Arizona Attorney General’s Office. You can reach the consumer fraud unit in Phoenix at 602-542-5763, in Tucson at 520-628-6504, or outside the Phoenix and Tucson metro areas at 1-800-352-8431. For suspected financial exploitation of an older adult, the Attorney General’s senior line is (844) 894-4735 toll-free or (602) 542-2124. Complaints can also be filed online at azag.gov.
You may also want to contact your bank, place a fraud alert with the credit bureaus, and let a trusted family member know what happened.
The Bottom Line
AI chatbots can be helpful companions for answering questions and getting everyday tasks done. But they are not private diaries, and they are not confidential advisors. The companies behind them may store your words, use them to train future versions, and in some cases let human reviewers read them. The safest approach is simple: ask general questions, turn off training in every chatbot you use, and keep your personal information out of the chat box entirely.
Sources:
Stanford HAI: Be Careful What You Tell Your AI Chatbot
Unite.AI: What the EU AI Act Doesn’t Tell Founders About Training Data
MemX: What AI Does With Your Chats
MemX: Opt Out of AI Training
Google Gemini Apps Privacy Notice
Google Support: Gemini Apps Activity
Redact: Turn Off Gemini Apps Activity
AI Toolbox: How to Turn Off Gemini
NHI Mgmt Group: Mainstream AI Chatbots Train on User Prompts by Default
Analytics Vidhya: AI Chatbot Privacy Guide
ConductAtlas: OpenAI Training Opt-Out
Microsoft Support: Copilot Privacy Controls
Big Technology: Check Your Chatbot’s Privacy Settings
X Help Center: About Grok
Redact: Stop Grok Training on Your X Posts
Offlist: Opt Out of AI Training Data
mePrism: Opt Out of Meta AI Training
Proton: Turn Off Meta AI
KTAR: Arizona AG Consumer Protection
Arizona Attorney General: Seniors
Arizona Attorney General: Consumer Protection Week