Last reviewed 29 Sept 2026
OpenAI
ChatGPT, Sora and the OpenAI API
OpenAI can access everything you type into ChatGPT, and by default uses consumer conversations to train its models. You can switch training off, and business and API customers are excluded by default. But chats are not encrypted from OpenAI, a court ordered it to keep chat logs, including deleted ones, and to hand over 20 million de-identified chats in a copyright lawsuit, and since February 2026 ChatGPT has shown ads on its Free and Go plans in the US and a growing number of other countries. Many people tell ChatGPT things they would never post publicly, so this matters more than it might seem.
- Headquarters
- San Francisco, California
- Jurisdiction
- United States
- Website
- openai.com
Your data is part of how it makes money.
What this means for you
- Treat ChatGPT like an email to a company, not a diary. Authorised OpenAI staff and service providers can access conversations in some situations.
- Turn off "Improve the model for everyone" in Data Controls if you don't want your chats used for training, or use Temporary Chat. Giving a thumbs up or down can still send that whole conversation for training.
- Deleting a chat is not always final. In 2025 a court ordered OpenAI to keep deleted chats as evidence in a lawsuit.
- Never paste passwords, medical records, ID documents or other people's private information into a chatbot.
Better alternatives
More private options for the same job.
Score breakdown
Apps & products
ChatGPT
Same as the companyConsumer plans (Free, Go, Plus, Pro) train on chats by default unless you opt out. Free and Go plans show ads in the US and a growing list of other countries.
ChatGPT Business & Enterprise
Better than the company overallNot used for training by default, with admin-controlled retention.
OpenAI API
Better than the company overallAPI data is not used for training by default.
The evidence
Every fact that feeds the score, newest first.
Can they read your data?
Whether your content is end-to-end or zero-access encrypted, so the company itself can't read it.
ChatGPT conversations are not end-to-end encrypted, and authorised OpenAI staff and service providers can access them, for example to investigate abuse or improve models.
Sources (4)
- Privacy policyOpenAI · Primary sourcearchived
- How OpenAI handles data in consumer servicesOpenAI Help Center · Primary source
- OpenAI Announces Plans for "Client-Side Encryption" for ChatGPTPrivacy Guides · 15 Nov 2025 · Reporting
- Fighting the New York Times' invasion of user privacyOpenAI · 12 Nov 2025 · Primary source
Business model
How the company makes money. Advertising and data-sharing models reward collecting more about you.
OpenAI shows ads in ChatGPT to logged-in adult users on its Free and Go plans, starting in the US in February 2026 and since expanded to other countries.
Ads are chosen by matching advertisers' ads with the topic of your conversation, your past chats and past ad interactions. OpenAI says advertisers don't get access to conversations and only receive aggregate performance data. Plus, Pro, Business, Enterprise and Education plans have no ads, and Free users can opt out of ads in exchange for fewer free messages.
Sources (3)
- Our approach to advertising and expanding access to ChatGPTOpenAI · 16 Jan 2026 · Primary sourcearchived
- Testing ads in ChatGPTOpenAI · 9 Feb 2026 · Primary source
- OpenAI Begins Testing Ads In ChatGPT For Free And Go UsersSearch Engine Journal · 9 Feb 2026 · Reporting
OpenAI says it keeps conversations private from advertisers and does not sell personal data.
OpenAI's privacy policy says it may share limited data with marketing partners to advertise OpenAI's own products elsewhere ('cross-context behavioral advertising'), which you can opt out of, and that it may receive information from advertisers, such as purchases you make with them.
Sources (4)
- Our approach to advertising and expanding access to ChatGPTOpenAI · 16 Jan 2026 · Primary sourcearchived
- Testing ads in ChatGPTOpenAI · 9 Feb 2026 · Primary source
- How OpenAI handles data in consumer servicesOpenAI Help Center · Primary source
- Privacy policy (updated 10 September 2026)OpenAI · Primary source
Track record
Breaches, scandals and regulator fines, and how the company responded.
OpenAI disclosed that AI agents in its research environment posted 53 user-provided images from training and evaluation data to third-party image-hosting sites as unlisted links.
Sources (2)
- Unsecured OpenAI agents posted 53 user images on the internet without the lab's knowledgeTechCrunch · 25 Sept 2026 · Reporting
- OpenAI's AI agents accidentally uploaded user-provided images to third-party sitesBleepingComputer · 26 Sept 2026 · Reporting
A breach at OpenAI's former analytics vendor Mixpanel exposed names, email addresses and rough locations of some API users, and of a limited number of ChatGPT users who had submitted help tickets or used the API platform website.
OpenAI says no chats, API requests, passwords or payment details were exposed.
OpenAI stopped using Mixpanel.
Sources (2)
- What to know about a recent Mixpanel security incidentOpenAI · 26 Nov 2025 · Primary sourcearchived
- OpenAI discloses API customer data breach via Mixpanel vendor hackBleepingComputer · 27 Nov 2025 · Reporting
Thousands of shared ChatGPT conversations, some with personal details, showed up in Google search results because of an opt-in "make this chat discoverable" box when creating a share link.
OpenAI removed the feature and asked search engines to remove indexed chats.
Sources (3)
- OpenAI is removing ChatGPT conversations from GoogleEngadget · 1 Aug 2025 · Reportingarchived
- Exclusive: Google is indexing ChatGPT conversations, potentially exposing sensitive user dataFast Company · Reporting
- OpenAI Is Pulling Shared ChatGPT Chats From Google SearchSearch Engine Journal · Reporting
Italy's data protection authority temporarily banned ChatGPT in 2023 over its legal basis for using personal data, missing information for users and the lack of age checks.
A related €15 million fine issued in 2024 was annulled by the Court of Rome in March 2026 on jurisdictional grounds.
The ban was lifted after OpenAI added privacy information, an opt-out and age checks for Italy.
Sources (2)
- Provvedimento del 30 marzo 2023Garante per la protezione dei dati personali · 30 Mar 2023 · Primary source
- OpenAI Prevails in Landmark Italian AI and GDPR Enforcement CaseWilson Sonsini · Reporting
A bug let some ChatGPT users see other users' chat titles, and may have exposed partial payment information (name, email, payment address, card type, last four digits and expiry date) of 1.2% of ChatGPT Plus subscribers active during a nine-hour window.
Full card numbers were not exposed, and OpenAI says the number of users whose data was actually seen by someone else was 'extremely low'.
OpenAI took ChatGPT offline, fixed the bug and notified affected users.
Sources (2)
- March 20 ChatGPT outage: Here's what happenedOpenAI · 24 Mar 2023 · Primary sourcearchived
- OpenAI: ChatGPT payment data leak caused by open-source bugBleepingComputer · 24 Mar 2023 · Reporting
Government & legal requests
What the company hands over to police, courts and governments, and whether it pushes back or reports on it.
OpenAI's transparency report shows it handed over user content in response to 62 of 75 government requests for content (84 accounts) from July to December 2025.
Sources (1)
- Government requests for user data, July–December 2025OpenAI · Primary source
A US court ordered OpenAI to hand over a sample of 20 million de-identified ChatGPT conversation logs to The New York Times and other news publishers suing it, under a protective order.
The logs are a random sample of consumer ChatGPT conversations from December 2022 to November 2024. OpenAI opposed the order; a district judge upheld it in January 2026. Business, Enterprise, Edu and API customers were not included.
Sources (4)
- OpenAI ordered to turn over 20 million ChatGPT chats to the New York TimesThe Decoder · 12 Nov 2025 · Reportingarchived
- In re OpenAI Copyright Infringement Litigation, 25-md-3143, Order (Dkt. 1087) affirming Magistrate Judge Wang's orders of Nov 7 and Dec 5, 2025US District Court, S.D.N.Y. (copy hosted by Ars Technica) · 5 Jan 2026 · Primary source
- Fighting the New York Times' invasion of user privacyOpenAI · 12 Nov 2025 · Primary source
- ChatGPT creator must turn over 20M chat logs in copyright litigation, federal judge saysABA Journal · 8 Jan 2026 · Reporting
A US court ordered OpenAI to preserve ChatGPT and API output logs that would otherwise have been deleted, including chats users had deleted, as evidence in The New York Times' copyright lawsuit.
The obligation to keep new data ended on 26 September 2025. Logs already kept under the order, and data from accounts flagged by the Times, are still preserved.
Sources (3)
- How we're responding to The New York Times' data demands in order to protect user privacyOpenAI · Primary sourcearchived
- OpenAI no longer has to preserve all of its ChatGPT data, with some exceptionsEngadget · 11 Oct 2025 · Reporting
- US Court Orders OpenAI to Preserve All ChatGPT Logs IndefinitelyCyberInsider · 9 Jun 2025 · Reporting
Data collection
How much the company collects, including tracking outside its own products.
OpenAI collects the content of your chats, uploaded files, images and voice, plus account, device and usage information.
Sources (1)
- Privacy policyOpenAI · Primary sourcearchived
AI & your conversations
Whether your chats, posts and files are used to train AI, reviewed by humans, or kept after you delete them.
ChatGPT uses consumer conversations to train OpenAI's models by default.
Sources (3)
- Data Controls FAQOpenAI Help Center · Primary sourcearchived
- How to stop AI companies from training on your dataEngadget · 28 Sept 2026 · Reporting
- How OpenAI handles data in consumer servicesOpenAI Help Center · Primary source
You can opt out of training in Data Controls, or use Temporary Chat, which isn't used for training and is kept for up to 30 days for safety purposes.
Sources (3)
- Temporary Chat FAQOpenAI Help Center · Primary sourcearchived
- Data controls in ChatGPTOpenAI Help Center · Primary source
- How to stop AI companies from training on your dataEngadget · 28 Sept 2026 · Reporting
Business, Enterprise and API data is not used for training by default.
Sources (2)
- Enterprise privacy at OpenAIOpenAI · Primary sourcearchived
- How OpenAI handles data in consumer servicesOpenAI Help Center · Primary source
Your control
Can you see, export and delete your data, and opt out without dark patterns?
You can export your ChatGPT data and delete individual chats or your whole account.
Sources (2)
- Data Controls FAQOpenAI Help Center · Primary sourcearchived
- How OpenAI handles data in consumer servicesOpenAI Help Center · Primary source
Jurisdiction
Which country's laws apply, and how much those laws protect you from surveillance.
As a US company, OpenAI can be compelled under US law, including the CLOUD Act, to hand over data it holds, wherever it is stored.
Sources (4)
- CLOUD Act ResourcesUS Department of Justice · Primary sourcearchived
- Promoting Public Safety, Privacy, and the Rule of Law Around the World: The Purpose and Impact of the CLOUD Act (white paper)US Department of Justice · Primary source
- OpenAI Report on Government Requests for User Data, July - December 2025OpenAI · Primary source
- OpenAI Government User Data Request Policy (effective 1 January 2026)OpenAI · Primary source
First published 3 Nov 2026