Someone emails you a link to a new AI tool. You click, type a question, get a useful answer. Then you wonder: where did that conversation go? Who can read it? Is it training the model on what you just wrote?
These are the right questions. The answers are less alarming than the headlines suggest, but more nuanced than "it's fine, don't worry." ChatGPT is safe for the things beginners use it for. It is not private in the way a sealed letter is private.
Here is a framework for using ChatGPT without giving away anything you would rather keep to yourself.
What ChatGPT stores
When you type something into ChatGPT, that text goes to OpenAI's servers. The conversation is saved to your account history by default. You can see it when you open the sidebar and look at past chats.
OpenAI encrypts data in transit and at rest (TLS 1.2+ and AES-256, according to their published security documentation). Your conversations are protected from outside interception. But encryption is not the same as privacy. OpenAI personnel and authorized service providers can access conversation data under certain circumstances, such as safety reviews or abuse investigations.
If you delete a conversation, OpenAI says it is typically removed from their systems within about 30 days, with exceptions for legal or security reasons.
Here is the trade-off: your conversations live on someone else's servers. That is not unique to ChatGPT. Gmail, Google Docs, Slack, and every other cloud service works the same way. The difference is that ChatGPT may also use your conversations to improve its models.
The training data question
On Free and Plus plans, OpenAI may use your conversations to train its models. This is on by default. You can turn it off.
Here is how:
- Open Settings in ChatGPT (click your name, then Settings)
- Go to Data Controls
- Toggle off "Improve the model for everyone"
Once that is off, your conversations are not used for model training. The toggle applies going forward. It does not retroactively remove your past conversations from training data.
If you are on a Team or Enterprise plan, your data is excluded from training by default. No toggle needed. API customers can also request Zero Data Retention, which means OpenAI does not keep prompts or responses after processing them.
We consider this the single most important setting for anyone concerned about privacy. If you do nothing else after reading this post, turn off that toggle.
Low stakes vs high stakes: what to share
Not all conversations carry the same risk. A simple framework helps here.
Low stakes: brainstorming ideas, drafting a blog post outline, asking for recipe suggestions, summarizing an article you pasted in, getting help with a code snippet, practicing a language. If a stranger read this conversation, you would not care. Share freely. This is what ChatGPT is good at, and the privacy cost is minimal.
Medium stakes: work documents that are not confidential, project plans, meeting notes, drafts of emails you plan to send. These are not secrets, but you might prefer they not be public. Using ChatGPT here is reasonable. If you have turned off the training toggle, the privacy risk is low.
High stakes: Social Security numbers, passwords, bank account numbers, medical records, confidential client information, trade secrets, anything covered by a nondisclosure agreement, anything that could identify a minor. Do not type these into ChatGPT. Not because ChatGPT is malicious, but because you are sending sensitive data to a third party's servers, and no setting change makes that fully private. If a lawyer types client information into ChatGPT, it can waive attorney-client privilege because the information has been disclosed to a third party.
The framework is simple. Before you type, ask: would we be comfortable if this text appeared on a server we do not control? If the answer is no, do not type it.
Settings worth checking
Beyond the training toggle, ChatGPT has a few settings that affect your privacy. None of these are complicated. We recommend spending two minutes going through them.
Chat history. Enabled by default. When on, conversations are saved to your account and appear in the sidebar. You can turn this off, which means conversations are not saved to your history. Turning off chat history does not prevent OpenAI from retaining conversations for 30 days for abuse monitoring.
Temporary Chat. Starts a conversation that is not saved to your history, not used for memory, and not used for training. The conversation still goes to OpenAI's servers and may be retained for up to 30 days for safety purposes. Think of it as incognito mode for ChatGPT. It reduces the footprint of your conversation but does not eliminate it.
Saved memories. ChatGPT can remember facts about you across conversations (your name, your job, preferences you have shared). You can turn this off in Settings under Personalization. You can also delete specific memories or clear all of them.
Shared links. If you create a public link to a conversation, anyone with the link can read it. Useful for sharing interesting exchanges, but the content is public. Do not share links to conversations that contain sensitive information.
Data export. You can request a copy of all your ChatGPT data through Settings. This includes your conversations and uploaded files. If you want to see what OpenAI has on you, this is how.
What ChatGPT gets wrong (and why that matters for safety)
There is a different kind of safety concern that has nothing to do with data privacy. ChatGPT can be confidently wrong. It can state incorrect facts with full assurance. It can cite sources that do not exist. It can give medical advice that is inaccurate. It can produce legal information that is wrong.
This does not make ChatGPT unsafe to use. It makes it unsafe to rely on without verification for anything that matters. If you are using it to brainstorm, draft, or explore ideas, the risk of error is low because you are reviewing the output with your own judgment. If you are using it to make a medical decision, a legal decision, or a financial decision without verifying the output independently, the risk is high.
We covered this in our guide to AI literacy, where we described the skill of knowing when to trust AI output and when to verify it. The same principle applies here. Proportionate verification for proportionate stakes.
For a practical walkthrough of how to use ChatGPT in low-risk ways, our starter guide for using ChatGPT at work covers the three behaviors that unlock most of the value without requiring you to share anything sensitive.
The honest summary
ChatGPT is safe for everyday use if you follow three rules:
- Turn off model training in your Data Controls settings
- Do not share high-stakes personal or professional information
- Verify output before acting on it for anything that matters
None of this requires technical knowledge. The settings take two minutes to configure. The framework takes one sentence to remember: if you would not want it on a server you do not control, do not type it.
For more on what ChatGPT can do (and where its limits are), see our guide to ChatGPT use cases.
FAQ
Is ChatGPT safe to use for personal tasks?
Yes, for everyday personal tasks like brainstorming, drafting, summarizing, and learning. ChatGPT encrypts your data in transit and at rest. The main privacy concern is that conversations may be used for model training on Free and Plus plans unless you disable that in Settings. As long as you are not entering sensitive personal information like Social Security numbers or medical records, the risk is low.
Does ChatGPT share my data with third parties?
OpenAI does not sell your personal data. According to their privacy policy, they may share information with service providers who help operate the service under confidentiality obligations, and they may disclose data when legally required. Authorized OpenAI personnel can review conversations for safety and abuse investigations. This is standard for cloud services.
Can my employer see what I type into ChatGPT?
If you are using a personal ChatGPT account on your own device, your employer cannot see your conversations directly. If you are using a company-provided ChatGPT Team or Enterprise account, your company's administrators may have access to usage logs and conversation data. Check your company's AI usage policy before using work accounts for anything personal.
How do I delete my ChatGPT data?
You can delete individual conversations by clicking the trash icon next to each chat in the sidebar. You can delete all conversations at once in Settings. You can also request a full data export and account deletion through OpenAI's privacy controls. Deleted conversations are typically removed from OpenAI's systems within about 30 days.
Is ChatGPT safe for kids to use?
ChatGPT requires users to be at least 13 (or older in some regions) and requires parental permission for users under 18. The tool is not designed for children. While it can be educational, it can also produce inappropriate content, and children may not have the judgment to verify inaccurate output. Parents should supervise use and consider whether the child understands the privacy implications of what they share.
Want to go further?
Spend two minutes in your ChatGPT settings today. Turn off model training. Review your chat history. Delete conversations you no longer need. That is the baseline.
If you want a structured starting point for your AI journey, take the free GWT survey at greatwithtech.com. It takes 2 minutes and maps your AI readiness across practical skills.