The questions people bring to ChatGPT are rarely the ones they’d talk about with someone close. A lump that showed up last month, a son who stopped calling, a pension that no longer covers the month. Those things get typed into a blank box because the box feels like nobody, and nobody can gossip.
But every one of those conversations travels to a company computer that already has your name, your email, and the card you paid with. Some of what you type is used to build the next version of the program. The rest sits in storage for a while after you think you’ve erased it.
In May of 2025 a federal judge ordered the company to hold onto records it would otherwise have deleted, and that included conversations people had already thrown away.
That order has since been lifted, and the program is still worth using. The trouble is that the settings meant to protect you arrive switched the wrong way, and they’re buried where nobody thinks to look. Everything below takes about ten minutes to fix.
First Things First – Why You Should Do This ASAP…
Nothing bad happens on the day you type something private into that box, so there’s no bill and no knock at the door to tell you anything went wrong. But the trouble shows up years later, and by then it arrives in a form you don’t get to argue with.
Here’s why you should do this as soon as possible:
- Because a company policy is a promise and not a law, so it can be rewritten on a Tuesday morning without anyone calling to tell you.
- Because everything you’ve typed up to today is already in storage, which leaves you only the conversations you haven’t had yet.
- Because a record of you rarely stays where it was made. Companies get bought, data gets sold, and a court can demand whatever’s left.
- Because the file gets more valuable the longer it runs. One question about a heart medication means nothing on its own, but six years of them is a medical history you never agreed to hand over.
- Because it’s all under one name and one payment card, so what you asked about your income sits in the same place as what you asked about your marriage and your firearms.
Daniel Cross, the author of Survival Before the Silence, keeps pointing readers back to one passage, and it’s Revelation 13:16-17:
“And he causeth all, both small and great, rich and poor, free and bond, to receive a mark in their right hand, or in their foreheads: And that no man might buy or sell, save he that had the mark, or the name of the beast, or the number of his name.”
What he draws out of those verses is that nobody gets arrested. Everyone simply gets identified, and after that the beast decides whether you’re allowed to buy food or not. A file did the work that soldiers used to do. Cross was writing about that years before anyone wanted to hear it, back when reviewers wrote him off as a fearmonger, and prepper circles have caught up with him since.
Many people saw this coming, and there’s no path back to how things were. So what’s left is to prevent what you still can, then adapt to the rest.
Find the Settings Menu
Everything in this article lives behind one menu, and getting to it differs depending on what you’re using.
- On a computer: go to chatgpt.com, click your profile picture or initial in the top right corner, then click Settings.
- On an iPhone or Android phone: open the app, tap your profile picture in the top right (on some versions, the three lines in the top left, then your name at the bottom), then tap Settings.
The two menus contain the same options but arrange them a little differently, so if something isn’t where this article says it is, scroll. The labels stay the same even when the order moves around.
Turn Off the Switch That Feeds Your Words Into the System
On the free, Plus, and Pro personal accounts, your conversations are used to train future versions of the program unless you say otherwise. Turning that off is the single most useful thing you can do, and it takes about fifteen seconds.
- Open Settings
- Click Data Controls
- Find the line that reads Improve the model for everyone
- Switch the toggle off
- Close the menu, reopen Data Controls, and confirm it’s still off
That last step is worth doing. Toggles occasionally fail to save, and there’s no confirmation message to tell you either way.
The change applies to your whole account and every device you sign in on, so you won’t need to repeat it on your phone. It only affects conversations from that moment forward, though, because anything already absorbed into a finished version of the program cannot be pulled back out.
There’s also a second route if the toggle gives you trouble. The company runs a privacy portal at privacy.openai.com where you can select Do not train on my content, and choosing it there updates the setting on your account as well. You don’t need to do both.
One habit undoes all of it. Under every answer sits a thumbs up and a thumbs down, and clicking either one permits to use that entire conversation for training regardless of your setting. If you’ve turned training off, leave those buttons alone.
Use Temporary Chat for Anything Sensitive
There’s a mode that leaves no trail behind, and most people have never noticed the button.
On a computer, look at the top right of the chat screen for a small icon that looks like a speech bubble with a dashed or dotted outline. On the phone app, start a new chat and tap the model name at the top, and you’ll find Temporary Chat in the list that drops down. The screen changes color slightly and tells you the chat won’t be saved.
What that actually means:
- The conversation never appears in your chat history
- Nothing from it is added to the program’s memory
- It isn’t used for training, even if your training toggle is on
- The company still holds a copy for up to 30 days before deleting it
But keep in mind that temporary isn’t the same as invisible, so it’s the right tool for a question about your prescription or your mortgage, not for anything you’d genuinely need to keep off a server.
So, ordinary questions go in a normal chat, and anything involving your health, your money, your family, or your firearms goes in a temporary one.
Read and Clear What It Remembers About You
ChatGPT keeps notes on you between conversations, so something mentioned once in passing can resurface months later in a completely unrelated chat. Mention that you’re diabetic, that you keep silver in the house, or that your daughter is going through a divorce, and it may write that down.
To see the list:
- Open Settings
- Click Personalization
- Click Manage Memories (sometimes listed as Manage Memory)
A window opens with everything it has recorded about you, line by line. Each entry has a small trash can or X beside it, and clicking that deletes just that one item. There’s usually a Clear All button at the bottom if you’d rather wipe the whole thing at once.
In the same Personalization menu you’ll find two separate switches worth understanding, because turning off one doesn’t turn off the other:
- Reference saved memories controls those written notes
- Reference chat history lets the program look back through your old conversations even when it hasn’t saved a note
Turning both off gives you a program that starts fresh every time. That’s less convenient, and it’s also the only way to be sure a detail from March doesn’t turn up in November. If you’d rather keep the feature, put a reminder on the calendar to read that list once a month.
Shut Off Old Shared Links
Sharing a conversation doesn’t send someone a copy. Instead, it creates a live web page that anyone holding the address can open, and thousands of those pages have been found through search engines over the years because the addresses got passed around or indexed.
- Open Settings
- Click Data Controls
- Find Shared Links and click Manage
- Delete the ones you no longer need, or use the option to delete all
Anyone who clicks an old address after that gets nothing. It’s worth checking even if you don’t remember sharing anything, since the button sits close to the copy button and gets hit by accident.
What You Should Never Type In, Regardless of Settings
No setting protects you from what you put in the box yourself. These belong somewhere else entirely:
- Your full Social Security number, or anyone else’s
- Bank account, credit card, or routing numbers
- Passwords, PINs, and the answers to your security questions
- Photographs of documents that show your address, account numbers, or signature
- Your exact home address, especially alongside anything about what you keep there
- Medical record numbers, insurance ID numbers, and dates of birth in full
Other people deserve the same care. A neighbor’s illness, a relative’s paperwork, a friend’s legal trouble – none of them agreed to go into a company’s records, and a conversation about someone is still a record of them.
There’s a legal point underneath all this that catches people off guard. What you tell a doctor or a lawyer is protected by privilege, and what you tell a chat program is not. If a court asks for it, you have nothing to invoke.
What Deleting Actually Does
Deleting a chat takes it off your screen right away, but the company doesn’t remove it from its own systems until up to 30 days later. To delete one conversation, hover over it in the sidebar, click the three dots, and choose Delete. To clear everything at once, go to Settings, then Data Controls, then Delete all chats.
Before you wipe anything, you may want a copy for yourself. Under Data Controls there’s an Export data option that emails you a downloadable file of every conversation you’ve had. It arrives as a folder you open in your web browser, and it’s worth doing once just to see the volume of what’s accumulated.
That 30-day gap has been pried open before. The order forcing the company to preserve everything ended on September 26, 2025, and normal deletion resumed, though a limited batch of records from April through September of that year is still held in storage because the newspaper suing them asked for it.
The lesson outlives the lawsuit. A deletion policy is a promise a company makes, and a judge can set that promise aside overnight without asking you or them.
Keep Your Name Off the Account
Two small choices give you a layer of separation that no toggle can provide.
Sign up with an email address that isn’t your full legal name, ideally a free one made for this purpose alone. Put a nickname or initials in the profile field rather than the name on your driver’s license. Neither makes you anonymous, since the payment method on a paid account carries your real identity, but both keep your name out of the conversation records themselves.
Run One on Your Own Computer
If you want the help without sending your words anywhere at all, you can run one of these programs on your own machine.
Free software called LM Studio is the easier of the two starting points, because it installs like any ordinary program and gives you a list of models to download with a click. Ollama does the same work with less hand-holding. Either way you download the program once, and after that it answers you with the internet unplugged.
What you need and what you give up:
- A computer made in roughly the last five years, with 16GB of memory or more
- Between 4 and 40 gigabytes of hard drive space, depending on which model you pick
- Answers that aren’t as sharp as the big paid services deliver
- No conversation ever leaving the house, and nothing to delete afterward
Start with one of the smaller models, usually labeled 7B or 8B, since those run on ordinary hardware without complaint. It’s the same reasoning that put a generator in the garage, because something you own outright beats something you rent by the month.
Make Sure You Hide Your Location
You never have to type your town for the program to know it, since every device you sign in from carries an IP address, and that alone puts you in a city. The phone app can also ask for location permission outright, the way a weather app does, and that box gets tapped without much thought.
So when you ask about anything nearby, the answer comes back looking like Google Maps, with distances measured from where you’re standing. It isn’t reading your Maps account, though. Your own phone handed over the starting point. That’s why it’s worth knowing how to stay off ChatGPT’s radar.
The good news is that going dark on it takes two minutes, and neither move costs you anything you’d miss.
The Ten-Minute Version
If you do nothing else, do these four things in one sitting:
- Settings, Data Controls, turn off Improve the model for everyone
- Settings, Personalization, open Manage Memories and clear what shouldn’t be there
- Settings, Data Controls, Shared Links, delete the old ones
- Learn where the Temporary Chat button is, and use it for anything personal
It takes only 10 minutes and the program stops keeping a file on you and goes back to being what most people assumed it was on the day they signed up.
You’ll Also Need This…
Worth knowing what you’re actually talking to. AI is a prediction machine, trained on more text than any of us could read in a lifetime, guessing the next word so well it passes for company. It doesn’t think. It collects.
And it’s spreading into things that have nothing to do with a chat box. Your utility meter reports your usage hour by hour. Your doorbell knows who comes and goes. Your car logs where it’s been and your phone confirms it. All of that gets sorted by the same kind of machine, on somebody else’s terms, whether you ever open an app or not.
Which is the argument this AI defense system makes, and it’s the one that changed how I look at my own house. Every convenience I’d installed over the years was quietly keeping records on me.
Not one of them asked first. The program goes device by device and shows you what each one gives away, then what to run instead. Some of it I fixed in an afternoon. The rest changed what I’ll buy from here on.
So do the same for the rest of the house, because the AI defense system I am talking about goes device by device and shows you what each one is handing over.
You may also like:
How to Make “Frontier Penicillin”
Build a Backyard Root Cellar With a Plastic Barrel (VIDEO)
Read the full article here
