ai · OpenAI
You’re creating on OpenAI, which means…
OpenAI is the rare one that hands the output to you instead of taking a licence in it — the consumer terms contain no content licence at all. What it keeps is the right to train on what you type, until you go and switch that off.
OpenAI gives you a different contract depending on where you are. This page reads all 2; every clause below says which one it comes from.
What they own
Nothing you make — and they assign you their share of it: “you (a) retain your ownership rights in Input and (b) own the Output. We hereby assign to you all our right, title, and interest, if any, in and to Output.” That sentence runs toward you. What OpenAI keeps is the product itself: “We and our affiliates own all rights, title, and interest in and to the Services.”
What you own
Your prompts and the answers, minus three carve-outs that live in the second document. Spoken replies are cut out by name: “Any rights in Output assigned to you do not include ChatGPT Voice Output”, which is “for non-commercial use only”. Code output “may be subject to third party licenses, including, without limitation, open source licenses.” And anything an App fetches into your answer “is not owned by you or OpenAI”.
What you don’t
A say in training unless you ask for one. OpenAI “may use Content to provide, maintain, develop, and improve our Services,” and the paragraph that follows is an opt-out rather than a consent: “If you do not want us to use your Content to train our models, you can opt out”. You may also not control the account itself — sign up with a work address and your employer’s administrator “will be able to control your account, including being able to access Content”.
What they can do with it
In the consumer terms, less than any other platform in this set. There is no content licence: “perpetual”, “sublicensable” and “royalty-free” appear zero times in the Terms of Use, and so does the phrase “you grant”. What the Terms of Use do take, one paragraph above the Content section, is your feedback: “you agree that we may use it without restriction or compensation to you.” The grants over your work sit in the Service Terms: publish a GPT and you give OpenAI, and every user of it, an “irrevocable, royalty-free license”; share to the Sora feed and you “give OpenAI the right to reproduce, distribute, modify, display and perform it for the purpose of operating and promoting the Services.”
The clauses
The grant runs the other way
As between you and OpenAI, and to the extent permitted by applicable law, you (a) retain your ownership rights in Input and (b) own the Output. We hereby assign to you all our right, title, and interest, if any, in and to Output.
Similarity of content. Due to the nature of our Services and artificial intelligence generally, output may not be unique and other users may receive similar output from our Services. Our assignment above does not extend to other users’ output or any Third Party Output.
“We hereby assign to you”
This is the opposite direction from the clause this tool usually quotes. OpenAI is not taking rights in your Output — it is handing over whatever rights it might have had in it.
The hedge is “if any”
The assignment covers “all our right, title, and interest, if any”. Whether machine-generated output attracts copyright at all is unsettled, and the phrase transfers whatever exists without claiming that anything does.
There is no content licence here
“Perpetual”, “sublicensable” and “royalty-free” appear zero times in the Terms of Use, and so does the phrase “you grant”. The document’s only “irrevocab-” word is “irrevocably”, in the jury-trial waiver, which has nothing to do with your work.
Except your feedback
One paragraph above the Content section sits the exception: “We appreciate your feedback, and you agree that we may use it without restriction or compensation to you.” It runs from you to OpenAI, unrestricted and uncompensated — and because it sits above the definition that makes “Content” mean Input and Output, feedback typed into the product is plausibly Input.
What ownership doesn’t cover
It does not reach the product that made the Output — under Our IP rights, “We and our affiliates own all rights, title, and interest in and to the Services.” And the conduct list adds one limit on the Output itself: you may not “Use Output to develop models that compete with OpenAI.”
Training is on until you turn it off
OpenAI Terms of Use — Content: Our use of content; Opt out ↗
Our use of content. We may use Content to provide, maintain, develop, and improve our Services, comply with applicable law, enforce our terms and policies, and keep our Services safe. …
Opt out. If you do not want us to use your Content to train our models, you can opt out by following the instructions in this article … Please note that in some cases this may limit the ability of our Services to better address your specific use case.
This is the only place training appears
The word “train” appears once in the Terms of Use, in this sentence, and zero times in the Service Terms. Neither document says anything else about what the models learn from.
An opt-out, not a consent
The default is that your Content is used. The way out is a help-centre article rather than a term of the contract, and nothing here promises the setting will keep existing.
“Develop” is the wide word
The permitted uses are “provide, maintain, develop, and improve our Services”. Three of those describe running the product you are using. “Develop” is not defined, and it is the one that reaches beyond it.
API and business users are sent somewhere else
These consumer terms say the Business Terms “govern use of ChatGPT Enterprise, our APIs, and our other services for businesses and developers.” That is a third contract, and it is not among the documents quoted here — so nothing on this page tells you how OpenAI treats API data.
The second document takes pieces back out of what you own
OpenAI Service Terms — Section 8, Voice Conversations. Updated: June 12, 2026. ↗
Voice conversations is a ChatGPT feature that allows you to speak to our models and have them speak back (“ChatGPT Voice Output”). ChatGPT Voice Output is for non-commercial use only and may not be distributed or repackaged as a standalone audio recording or any other sound file. Any rights in Output assigned to you do not include ChatGPT Voice Output.
It overrides the ownership line
The Terms of Use say you own the Output. This sentence removes the spoken version from that assignment — and the Service Terms open by saying which one wins: “If there is a conflict between the Service Terms and your Agreement, the Service Terms will control.”
Non-commercial, and non-portable
Two separate restrictions in one sentence: you may not use it commercially, and you may not save it out as a standalone audio file. A podcast, an ad read or an audiobook built from ChatGPT Voice Output sits outside what this clause allows.
Two more carve-outs nearby
Section 4 says code output “may be subject to third party licenses, including, without limitation, open source licenses.” Section 7(b) says content an App retrieves into your answer “is not owned by you or OpenAI, and may be subject to the terms of the Application.”
And Content leaves on your instruction
Section 7(b) also runs outward: when you enable an App, “you authorize and instruct us to send your App Data to the applicable Application, which will be handled in accordance with its terms.” App Data is “applicable portions of your Content”, and OpenAI is “not responsible for App Data after it has been provided to an Application other than Applications owned by us.”
The one licence you actually grant
OpenAI Service Terms — Section 5, GPTs: Distribution and Promotion of GPTs ↗
By sharing your GPT with others, you grant a nonexclusive, worldwide, irrevocable, royalty-free license: (i) to OpenAI to use, test, store, copy, translate, display, modify, distribute, promote, and otherwise make available to other users all or any part of your GPT (including GPT Content); and (ii) to the extent Output from your GPT includes your GPT Content, to users of your GPT to use, store, copy, display, distribute, prepare derivative works of and otherwise use your GPT Content.
Once, in two documents
The phrase “you grant” appears exactly one time across the Terms of Use and the Service Terms, and this is it. Two other permissions run the same direction without the word: “you give OpenAI the right” in the Sora clause, and the Feedback line in the Terms of Use.
Irrevocable
The only appearance of “irrevocable” across both documents, and this is it — the nearest word in the Terms of Use is “irrevocably”, in the jury-trial waiver. Your GPT can be pulled from the store, but nothing in the clause lets you withdraw the licence you gave when you published it.
Two grants in one sentence
Part (i) runs to OpenAI. Part (ii) runs to every user of your GPT, and what they get includes the right to “prepare derivative works of” your GPT Content.
And it can be removed anyway
Subsection (d) sits four lines below: “We may reject or remove any GPT from our Services at any time for any reason without notice to you”.
Likeness is a rule you must follow, not a right you hand over
OpenAI Service Terms — Section 6, Image and Video Capabilities ↗
You may not use Visual Capabilities to assist in identifying a person nor to solicit or infer private or sensitive information about a person. You may not use Visual Capabilities to reproduce the likeness of any person without express consent and all necessary rights. By choosing to share an image or video publicly on the Services, such as by sharing it to the Sora feed or uploading it to create a Sora cameo, you represent you have all necessary rights, and you give OpenAI the right to reproduce, distribute, modify, display and perform it for the purpose of operating and promoting the Services. You also agree to give users the limited right to reproduce and remix that content solely on the Services. Uploading a cameo or sharing an image or video does not give other users any rights to use those materials outside of the Services.
The word “likeness” appears once
Once in the Service Terms, never in the Terms of Use — and where it appears it is a restriction on you, not a grant of your face to OpenAI. Clearing the person depicted is your obligation.
Sharing is what triggers it
The rights you hand over attach to the act of posting publicly — to the Sora feed, or as a cameo — not to using the feature. Generate privately and this sentence never fires.
Bounded, and then bounded again
OpenAI’s right is tied to “the purpose of operating and promoting the Services.” Other users’ right is “solely on the Services.” The clause then repeats the limit in plain words: sharing “does not give other users any rights to use those materials outside of the Services.” No platform elsewhere in this tool draws that line.
Your work email can move the account, and everything in it
OpenAI Terms of Use — Using our Services: Corporate domains ↗
If you create an account using an email address owned by an organization (for example, your employer), that account may be added to the organization's business account with us, in which case we will provide notice to you so that you can help facilitate the transfer of your account (unless your organization has already provided notice to you that it may monitor and control your account). Once your account is transferred, the organization’s administrator will be able to control your account, including being able to access Content (defined below) and restrict or remove your access to the account.
The address decides it
Which email you signed up with, not which plan you bought, is what can move the account — along with the Content inside it — under someone else’s control.
The notice has an exception
The promise to tell you carries a parenthetical: no notice is owed if your organisation has already said it may monitor and control your account. Many employment IT policies say exactly that.
And a different contract then applies
Once the account sits in a business workspace, the Service Terms section on Enterprise and Business governs, where Administrators “may be able to (a) access, share and remove Content; and (b) access logging and information about End Users’ use of ChatGPT Enterprise.”
The 2 agreements
Due Dilly reads the paperwork and reports what it says. This is not legal advice.