NewsCompaniesFlawsGuides
Get the app

Is ChatGPT safe?

From Is it safe?, in the HARD wiki · Updated
This article is about the ChatGPT AI chatbot. For other apps, see Is it safe?.

Safe to use with care: a mainstream, well-run service, but your chats can train its models and shape ads unless you change settings, so keep sensitive details out.

ChatGPT is run by OpenAI, a large US company with a public record of fixing problems and saying so. Its known incidents are a 2023 bug that showed some users other people’s chat titles and partial payment details, and a 2025 breach at an analytics supplier that did not include chats.

The real risk is what you type. By default, chats on Free, Plus and Pro can be used to train OpenAI’s models, and Free and Go users in some countries now see ads. Both can be switched off or reduced in settings.

ChatGPT is made by OpenAI, based in San Francisco. Its privacy policy says your data is processed and stored on servers in the United States and in countries where its affiliates and service providers operate.[1]

The real risks

What ChatGPT itself says it collects, what regulators and courts have found, its security record and the scams that borrow its name. Each line links to the page it comes from.

What it collects

From ChatGPT’s own privacy policy:

  • Account details: your name, contact details, login credentials, date of birth, payment information and transaction history.[1]
  • Everything you give it: prompts, files, images, audio and video, and data from services you connect, such as email or cloud drives.[1]
  • Your phone’s contacts, but only if you choose to connect them, so it can tell you which contacts also use OpenAI’s services.[1]
  • Device and usage data: IP address, browser and device type, device identifiers, time zone, features you use and the content you engage with.[1]
  • Your general location from your IP address, and precise GPS location only if you choose to share it.[1]
  • For Free and Go users: your ad history and interests, and information from advertisers and data partners, such as purchases you make from them.[1][2]
  • Information you give an age or identity check, such as a selfie through its verification partner, if you need to prove you are an adult.[1][3]

App store label. Apple’s listing shows only “Data Linked to You”: contact info, user content, search history, identifiers, usage data, diagnostics, coarse location, and health and fitness data. Usage and advertising data are listed for third-party advertising. No “Data Used to Track You” section is shown. Apple has not verified the label.[4]

Regulators and courts

Fines, orders, investigations and lawsuits, newest first. A lawsuit is what the side bringing it alleges until a court decides.

WhenWhat happened
Berlin’s data protection commissioner warned that personalised ChatGPT ads built from past and current chats carry serious privacy risks. In Europe, OpenAI plans to personalise ads only with your consent; other ads use the current chat only.[5]
A Rome court (ruling no. 4153/2026) upheld OpenAI’s appeal against Italy’s €15 million privacy fine. The Garante took its decision down from its website after the ruling.[6]
Lawsuits against OpenAI over mental-health harms, including the Raine family’s case over Adam Raine’s death, were put forward for joint handling as “ChatGPT Product Liability Cases”. OpenAI says a court has since joined them; it disputes the claims.[8][9][10]
A court order in The New York Times’ lawsuit had made OpenAI keep deleted and temporary chats. OpenAI says that duty ended on 26 September 2025, though it still holds some April–September 2025 data under legal hold.[11]
The US Federal Trade Commission ordered OpenAI and six other companies to explain how their chatbots affect children and teens, how they limit young users’ access, and how they use conversations. This is a study, not an enforcement action.[12]
Italy’s data protection authority, the Garante, fined OpenAI €15 million, saying it trained ChatGPT on personal data without a proper legal basis, failed to report the March 2023 breach and lacked age checks. A court later overturned it.[6]
The Garante ordered OpenAI to temporarily stop processing Italian users’ data, citing poor information to users, no clear legal basis for training, and no age verification.[13]

Breaches and security flaws

  • : OpenAI said an attacker took data from Mixpanel, an analytics supplier: names, email addresses, rough location and browser details of some API users and some ChatGPT users. No chats, passwords, payment details or API keys were exposed.[7]
  • : OpenAI said a bug let some users see other people’s chat titles, and that 1.2 per cent of active Plus subscribers may have had their name, email, billing address, card type, last four digits and expiry date shown to others.[14]

In HARD’s Situation Room: Fake OpenAI, SpaceX and Anthropic event invites steal Microsoft 365 and Google logins, even past two-factor (7 Nov 2026); ChatGPT: opening someone's shared chat could hand them your private ones, until a fix on 26 October (1 Nov 2026).

Changes that matter

  • : OpenAI began testing ads in ChatGPT in the United States for Free and Go users, later expanding to other countries. It says advertisers don’t see your chats and under-18 accounts get no ads.[2]

Scams that use ChatGPT

Fake OpenAI support accounts
Accounts on social media offer to help with your ChatGPT account and ask for your login. OpenAI says it never discusses account details over social media and you should not engage.[15]
Phishing emails that look like OpenAI
Messages ask you to confirm a payment or log in. OpenAI lists its real email domains and says it never asks for passwords, API keys or verification codes by email, text or chat.[15][7]
Copycat ChatGPT apps
Look-alike apps use the ChatGPT name. OpenAI links to its official apps from its help centre; on the App Store the genuine seller is OpenAI OpCo, LLC.[15][4]

What to do

In order, with the place to find each one in ChatGPT.

  1. Turn on multi-factor authentication.Settings → Security and login → Multi-factor authentication (MFA). Choose an authenticator app, push notifications, text message or a passkey, then follow the steps.[16]
  2. Stop your chats being used for training.Web: account menu → Settings → Data controls → turn off Improve the model for everyone. Mobile: sidebar → profile icon → Data controls. Giving thumbs-up or thumbs-down feedback can still send that whole chat for training.[17]
  3. Use Temporary Chat for anything sensitive.Open a new chat and select Temporary. These chats stay out of your history, don’t update memory, aren’t used for training and are deleted within 30 days.[18][17]
  4. Check what Memory has saved.Settings → Personalization → Memory. Review or delete saved memories, or turn Memory off. Turning it off doesn’t delete your past chats.[19]
  5. Delete old shared links.Settings → Data controls → Shared links → Manage. Delete links one by one or use Delete all shared links. Copies other people already saved stay in their accounts.[20]
  6. Limit ad personalisation (Free and Go plans).Settings → Ad Controls. Turn off ad personalisation and clear the data used for ads. You’ll still see ads based on the current chat.[2]
  7. Link your teen’s account with parental controls.Settings → Parental controls → Add family member, then send the invite. You can set quiet hours and turn off memory, voice, image creation and training. You can’t read their chats.[21][22]

When not to use it

  • Don’t type passwords, account numbers, health details or other sensitive information, especially into chats you might share.[20]
  • Don’t rely on it for facts that matter, such as medical, legal or money decisions, without checking another source. OpenAI says its answers may not be accurate.[1]
  • Children under 13 shouldn’t use it, and under-18s need a parent’s permission.[1]
  • On a work or school account, assume the administrator can see and control what you write.[1]

What HARD couldn’t verify

Claims you may see elsewhere that HARD couldn’t confirm from a primary source on 7 Nov 2026, so they aren’t in the record above.

  • News reports give further detail about what the Raine family and other families allege about ChatGPT; HARD did not confirm those details on a court page.
  • Claims about fake ChatGPT browser extensions stealing data come from security vendors’ blogs; HARD found no OpenAI or regulator page documenting a specific case.

Questions

Is ChatGPT safe to use?

For most people, yes, with care. OpenAI is a large, regulated company and its known incidents have been bugs and a supplier breach, not mass leaks of chats. Turn on multi-factor authentication, turn off training if you prefer, and keep sensitive details out of your chats.

Does ChatGPT use my chats to train its AI?

On Free, Plus and Pro, yes by default. You can turn this off under Settings → Data controls → Improve the model for everyone, or use Temporary Chat. Business, Enterprise and Edu workspaces are not used for training by default.

Does ChatGPT sell my data?

OpenAI says it doesn’t sell personal data and doesn’t share your chats with advertisers. Free and Go users may see ads chosen from the current chat and, if personalisation is on, other signals. In the US it may share limited data with marketing partners to advertise its own products, which you can switch off.

Is ChatGPT safe for kids?

It isn’t for children under 13, and under-18s need a parent’s permission. OpenAI estimates users’ ages and gives likely teens stricter content limits. Parents can link accounts to set quiet hours and get alerts about serious safety concerns, but can’t read their teen’s chats.

Are deleted ChatGPT chats really deleted?

OpenAI says deleted chats and Temporary Chats are removed from its systems within 30 days, unless it must keep them for legal or safety reasons. A court order made it keep them for part of 2025; that ended on 26 September 2025.

AI chatbots on HARD: Is Character.AI safe?, Is DeepSeek safe?. All of them: Is it safe?

References

Primary sources only: ChatGPT’s own pages, regulators, courts and app stores. Each was checked on 7 Nov 2026.

  1. OpenAI. US Privacy Policy, 10 Sep 2026. openai.com, checked 7 Nov 2026
  2. OpenAI Help Center. Ads in ChatGPT, 3 Nov 2026. help.openai.com, checked 7 Nov 2026
  3. OpenAI. Our approach to age prediction, 25 Aug 2026. openai.com, checked 7 Nov 2026
  4. Apple App Store. ChatGPT on the App Store. apps.apple.com, checked 7 Nov 2026
  5. Berliner Beauftragte für Datenschutz und Informationsfreiheit. Datenschutzbeauftragte sensibilisiert für personalisierte Werbefunktion bei ChatGPT, 2 Oct 2026. datenschutz-berlin.de, checked 7 Nov 2026
  6. Garante per la protezione dei dati personali (Italy). ChatGPT, il Garante privacy chiude l’istruttoria: sanzione di 15 milioni di euro, 20 Dec 2024. garanteprivacy.it, checked 7 Nov 2026
  7. OpenAI. What to know about a recent Mixpanel security incident, 19 Dec 2025. openai.com, checked 7 Nov 2026
  8. Judicial Council of California. Civil Case Coordination Proceeding (JCCP) Log, JCCP 5431 ChatGPT Product Liability Cases. courts.ca.gov, checked 7 Nov 2026
  9. OpenAI. Our approach to mental health-related litigation, 17 Dec 2025. openai.com, checked 7 Nov 2026
  10. OpenAI. An update on our mental health-related work, 27 Feb 2026. openai.com, checked 7 Nov 2026
  11. OpenAI. How we’re responding to The New York Times’ data demands in order to protect user privacy, 22 Oct 2025. openai.com, checked 7 Nov 2026
  12. US Federal Trade Commission. FTC Launches Inquiry into AI Chatbots Acting as Companions, 11 Sep 2025. ftc.gov, checked 7 Nov 2026
  13. Garante per la protezione dei dati personali (Italy). Intelligenza artificiale: il Garante blocca ChatGPT, 31 Mar 2023. garanteprivacy.it, checked 7 Nov 2026
  14. OpenAI. March 20 ChatGPT outage: Here’s what happened, 24 Mar 2023. openai.com, checked 7 Nov 2026
  15. OpenAI Help Center. Verifying Communications from OpenAI, 27 Oct 2026. help.openai.com, checked 7 Nov 2026
  16. OpenAI Help Center. Managing multi-factor authentication (MFA), 21 Oct 2026. help.openai.com, checked 7 Nov 2026
  17. OpenAI Help Center. Data controls in ChatGPT, 5 Nov 2026. help.openai.com, checked 7 Nov 2026
  18. OpenAI Help Center. Temporary chat in ChatGPT, 31 Oct 2026. help.openai.com, checked 7 Nov 2026
  19. OpenAI Help Center. Memory in ChatGPT, 31 Oct 2026. help.openai.com, checked 7 Nov 2026
  20. OpenAI Help Center. Sharing conversations and scheduled tasks in ChatGPT, 22 Oct 2026. help.openai.com, checked 7 Nov 2026
  21. OpenAI Help Center. Managing parental controls in ChatGPT, 3 Nov 2026. help.openai.com, checked 7 Nov 2026
  22. OpenAI. Introducing parental controls, 13 Jul 2026. openai.com, checked 7 Nov 2026