Is ChatGPT Safe to Use in 2026? Privacy, Data and Your Mac

Yes, ChatGPT is safe for everyday use if you download the official app and change a few settings. On Free, Go, Plus and Pro, OpenAI can use your chats to train its models until you turn off Improve the model for everyone, deleted chats are erased within 30 days, and nothing stops you from pasting something you'll regret. The risk is mostly what you type.

The worry usually shows up mid-task. I was halfway through pasting a client's contract into ChatGPT to ask for a plain-English summary when I stopped and wondered where that text actually goes. This guide answers that with OpenAI's own documentation: what ChatGPT collects, what trains the models, how long things are kept, what the Mac app can see on your screen, and which switches I'd flip before your next chat.

Is ChatGPT safe to use?

ChatGPT is safe in the sense most people mean: OpenAI doesn't sell your personal data according to its privacy policy, and the official app is a normal download from OpenAI, not a threat to your Mac. Where it isn't automatically safe is privacy. Your prompts, files and voice input are stored on OpenAI's servers, and on personal plans they can be used for training by default.

So I split the question in two. "Will ChatGPT hurt my computer?" is almost always no. "Should this particular piece of text leave my Mac?" depends on what it is, which plan you're on and how your settings are set. The rest of this article is about that second question.

What data does ChatGPT collect?

ChatGPT collects what you give it plus the usual technical trail. OpenAI's privacy policy lists these categories:

• User content: your prompts, uploaded files, images, audio and video, and data from services you connect.
• Account details: name, contact details, date of birth, payment information and transaction history.
• Log, usage and device data: IP address, browser, operating system, device identifiers, features you use and when.
• Location: a general area worked out from your IP address, and precise location only if you choose to share it.
• Contacts: your address book, only if you choose to connect it.

One line in the US version of the policy is easy to miss. Depending on your choices, OpenAI may share limited data with marketing partners to advertise its own products elsewhere, which some state laws call targeted advertising. You can opt out with the marketing privacy control in your account settings, or with Global Privacy Control in a browser that sends it, per the privacy policy.

Does ChatGPT use your conversations for training?

On personal plans, yes, unless you opt out. OpenAI's Data Controls FAQ says data sharing is on by default for Free, Plus and Pro accounts, and the pricing page lists the same default for Go with an opt-out. For Business, Enterprise, Edu and ChatGPT for Healthcare, OpenAI doesn't train on your content by default.

How to turn off training in ChatGPT on a Mac

The switch sits in the same spot in the browser and the Mac app, and it follows your account to every device once you're signed in:

1. Open ChatGPT and click your profile or account menu.
2. Choose Settings.
3. Open Data controls.
4. Turn off Improve the model for everyone.
5. Click Done.

Turning it off keeps your chat history; it doesn't delete or hide anything. If you use ChatGPT signed out, the choice is saved only in that browser. The Data Controls FAQ also notes two exceptions: Codex has its own setting for including environments, and sessions in OpenAI's Daybreak Trusted Access program can still be used for model improvement on a personal account even with the setting off.

How long does ChatGPT keep your chats?

Regular chats stay in your history until you delete them. Once you delete a chat or your account, OpenAI's privacy policy says it removes the data within 30 days, unless it has to keep it longer for legal reasons, fraud and abuse prevention or financial records.

Temporary Chats are the private-browsing option. They don't appear in your history, don't create or update memories, aren't used for training and are deleted within 30 days, according to the Data Controls FAQ. If you save a Temporary Chat, it becomes a normal chat and follows your usual settings.

The 30-day rule had a gap in 2025. A court order in the New York Times copyright case made OpenAI keep consumer chats that would otherwise have been deleted. OpenAI says that obligation ended on September 26, 2025, and that it still holds a limited set of user data from April to September 2025 under tight access for the case. I mention it because "deleted" can depend on events no settings page controls.

What ChatGPT memory knows about you

ChatGPT memory has two parts: saved memories, which are details it keeps on purpose, and chat history reference, which lets it draw on past conversations. Both are controlled in Settings > Personalization > Memory, per OpenAI's Memory FAQ.

Memory is useful, and it's also where things you mentioned once can resurface. I once asked ChatGPT for a packing list and it worked in a detail from a medical question I'd asked weeks earlier. Nothing leaked, but it reminded me that the app connects dots. A few details from the Memory FAQ are worth knowing:

• Deleting a chat doesn't necessarily delete a saved memory created from it. Remove the memory too.
• To clear everything at once, open the memory summary's ••• menu and choose Delete and turn off memory. Your chats stay.
• OpenAI may hold logs of deleted saved memories for a 30-day window, for safety and debugging.
• Before starting a Temporary Chat, you can pick Unpersonalized so it ignores your memories and custom instructions.

Is the ChatGPT Mac app safe?

The official ChatGPT app for Mac is safe to install when you get it from chatgpt.com/download, the address OpenAI's system requirements page points to. It requires macOS 14 and runs on Apple silicon and Intel Macs. Lookalike apps and "ChatGPT for Mac" downloads from other sites are where the real malware risk sits, so skip them.

The Mac app changed this year. Since July 9, 2026, the new desktop app combines Chat, Work and Codex, and the earlier Mac app has been renamed ChatGPT Classic and is still supported, per the same OpenAI page. Pressing Option-Space opens the Chat Bar, so ChatGPT is always one keystroke away. That's convenient, and it also makes pasting without thinking easier.

There's one bit of history Mac users should know. In July 2024, The Verge reported that the Mac app was storing conversations on disk as plain text that other apps could read. OpenAI shipped an update that encrypts them. Keep the app updated and that's behind you.

What macOS permissions ChatGPT asks for

ChatGPT for Mac only sees your screen or controls other apps after you grant macOS permissions. Two of them do the heavy lifting:

Computer Use is the one to think about. When it's on, ChatGPT can operate Mac apps for you. OpenAI's Computer Use docs say it asks before using each app unless you choose Always allow, may ask before sensitive actions, can't approve security or privacy prompts, and can't automate terminal apps. OpenAI also cautions that in your signed-in browser its actions carry your account's authority, and that web pages may contain misleading content meant to steer it.

An opt-in macOS-only option even lets it keep working after your Mac locks, through an Apple authorization plug-in. I leave that off. When I let an agent click around for me, I want to be sitting there.

How to check or remove ChatGPT's access on your Mac

1. Choose Apple menu > System Settings.
2. Click Privacy & Security in the sidebar.
3. Under Screen & System Audio Recording, switch off ChatGPT if you don't use screenshots (Apple's guide).
4. Go back and open Accessibility, then switch off ChatGPT or Codex Computer Use if you don't use Computer Use (Apple's guide).
5. In ChatGPT, open Settings > Computer use and delete any apps in the always-allowed list you no longer want it to touch (OpenAI).

Has ChatGPT ever been hacked or had a data breach?

In late 2025, OpenAI reported an incident that came through a vendor rather than its own systems. That November, an attacker got into Mixpanel, an analytics company OpenAI used on its API platform. OpenAI's incident report says names, email addresses, approximate location, browser and operating system details, and account IDs were exposed for API users and a limited number of ChatGPT users who had filed help tickets or signed in to platform.openai.com.

Chats, passwords, API keys and payment details weren't part of the exposed data, OpenAI says, and it removed Mixpanel from its products. The practical risk from leaked names and emails is phishing. OpenAI says it won't ask for your password, API keys or verification codes by email, text or chat, so any message that does is fake.

Lock down your ChatGPT account

1. Open Settings > Security and login.
2. Under Multi-factor authentication (MFA), pick an authenticator app, push notifications, a text code or a passkey (OpenAI).
3. While you're there, choose Log out of all devices if you've signed in somewhere you no longer use. It can take up to 30 minutes to take effect.

Turning on MFA doesn't sign out other sessions on its own, which is why I do both at once.

What you should never type into ChatGPT

Treat ChatGPT like a smart colleague at another company: helpful, but outside your walls. Even with training off, your chats sit on OpenAI's servers until you delete them, and OpenAI's terms of use say you shouldn't rely on the output as a sole source of truth. These are the things I keep out:

• Passwords, API keys, one-time codes and recovery phrases. No reason ever justifies pasting these.
• Government ID, bank and card numbers. Redact them from any document before you upload it.
• Health or legal details tied to a name. Ask the general question instead.
• Other people's private information. A colleague's performance review or a client's contract isn't yours to share with a third party.
• Anything under NDA or your employer's confidentiality rules. If your company has ChatGPT Business or Enterprise, use that workspace rather than your personal account.

That contract from the opening is a good test case. What I do now is strip the names, amounts and addresses, paste only the clause I don't understand, and ask what it means. I get the same answer without handing over the whole document.

Shared links, ads and connected apps

Three features change who can see your chats, and each has its own controls.

Shared links. A link from a personal account can be opened by anyone who has it, and it has no expiry date or per-person access. OpenAI's shared links FAQ says they aren't meant for search indexing but aren't private either, since links get forwarded. Review and delete them in Settings > Data controls > Shared links > Manage.

Ads. Since February 9, 2026, ChatGPT shows ads on the Free and Go plans; Plus, Pro, Business, Enterprise and Edu are ad-free, per OpenAI's ads FAQ. Advertisers get aggregated reporting such as views and clicks, not your chats, memories or personal details. Ads can still be picked using your current conversation. In Settings > Ad Controls you can turn off ad personalization, so other chats and your ad history aren't used, or clear your ads data. Temporary Chats show no ads.

Connected apps. Connect Gmail or a file service and ChatGPT can read what's in it, which becomes user content under the privacy policy. The Memory FAQ notes you may need to disconnect an app to fully remove information that came from it. Connect only the accounts a task needs.

A 10-minute ChatGPT privacy check for Mac users

If you do nothing else, run through this list once, then again whenever you change plans or install a new version of the app.

1. Confirm the app came from chatgpt.com/download and is up to date.
2. Turn off Improve the model for everyone in Settings > Data controls if you don't want your chats used for training.
3. Review saved memories in Settings > Personalization > Memory.
4. Delete old shared links in Data controls > Shared links.
5. Turn on MFA in Security and login.
6. On Free or Go, set Ad Controls the way you want them.
7. In System Settings > Privacy & Security, remove Screen Recording and Accessibility access you don't use.
8. Use a Temporary Chat for anything you wouldn't want in your history.

Not sure ChatGPT is the assistant you want to trust with this? Our ChatGPT alternatives page compares it with Claude, Gemini, Perplexity and Proton's Lumo on privacy and ads, and Gemini vs. ChatGPT goes deeper on those two. Once your settings are where you want them, these ChatGPT prompts are a good place to start.