Who owns your ChatGPT conversations? What 12 AI apps' terms say

The Morse teamPublished November 3, 20269 min read

You do, on paper. As of November 3, 2026, OpenAI's terms say that between you and OpenAI, you "retain your ownership rights in Input" and "own the Output," so your ChatGPT conversations are yours. Ownership doesn't stop OpenAI from keeping a copy and using it, though. The same terms let OpenAI use your content to "develop, and improve our Services" unless you opt out, and some apps take a license to your chats that never ends.

We read the consumer terms of service of 12 AI chat apps on November 3, 2026: ChatGPT, Claude, Gemini, Grok, Meta AI, Microsoft Copilot, DeepSeek, Mistral's Le Chat (now Vibe), Character.AI, Replika, Duck.ai, and Proton's Lumo. For each we wrote down who owns what you type, who owns what the AI writes, what license the company takes, whether deleting ends it, and what you're barred from doing with the output. Perplexity's terms page blocked our reader all day, so it isn't in the table. This isn't legal advice.

Most top results for this search answer a different question, who owns the company behind ChatGPT. The answer for your conversations is in each company's terms.

Who owns your ChatGPT conversations?

You own both halves, as far as OpenAI is concerned. OpenAI's terms of use (effective January 1, 2026) say that "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," and add: "We hereby assign to you all our right, title, and interest, if any, in and to Output" (OpenAI).

"As between you and OpenAI" means the promise binds OpenAI and no one else. "If any" means OpenAI hands over whatever rights it has, which may be few. The US Copyright Office concluded in January 2025 that "prompts alone do not provide sufficient human control to make users of an AI system the authors of the output" (Copyright Office, Part 2 report), and the Supreme Court declined on March 2, 2026 to hear Thaler v. Perlmutter, leaving in place the rule that a copyright needs a human author (Supreme Court docket 25-449).

OpenAI can still use its copy. Its terms say "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," and point you to a separate opt-out for training (OpenAI). We cover that switch in does ChatGPT train on your data.

Who owns your AI chats in each app?

Read on November 3, 2026, from each company's own consumer terms. "Input" is what you type or upload, and "output" is what the AI writes back.

None of the 12 claims to own what you type, and 10 say in writing that the output is yours or that the company claims none of it. Meta's and Replika's terms don't address the output. The apps differ on the license each one takes.

Proton's Lumo gets credit here. Its ownership and similarity clauses use OpenAI's words almost exactly, then add one OpenAI's don't have: "Company will not use any of your content to maintain, develop, or improve our Services. Your content is your own" (Proton). Duck.ai's license is nearly as narrow, "only to process your Prompts and provide you the Outputs" (DuckDuckGo).

What license do AI companies take to your chats?

Four of the 12 take licenses with no end date or few limits.

• Grok. By submitting inputs, you give xAI "an irrevocable, perpetual, transferable, sublicensable, royalty-free, and worldwide right" to use, copy, modify, publish, and "make derivative works of" them "for any purpose" (xAI). If you use Grok without logging in, xAI's terms say you grant "full rights to use any data you provide to or obtain from our Service for product development and model training purposes." We cover Grok's training settings in is Grok private.
• Character.AI. The license covers what you submit and what you generate, and includes the right to "exploit, commercialize and otherwise use" it, to let other users "remix" it, and to "pass these rights along to" companies Character.AI has contracts with (Character.AI).
• Replika. Uploading content grants Replika and its affiliates a "perpetual, irrevocable license" to use it to run the service and for "the promotion, advertising or marketing thereof." The next sentence limits that: Replika "never" uses or discloses "the content of your Replika conversations for marketing or advertising purposes" (Replika).
• Microsoft Copilot. Microsoft may "copy, distribute, transmit, publicly display, publicly perform, edit, translate, and reformat" your prompts and responses, and says "we don't have to pay you, ask your permission, or tell you when we do" (Microsoft).

OpenAI, Anthropic, and DeepSeek grant themselves no license by that name. Their terms say what they "may use" your content for instead, which comes to much the same thing. Anthropic's adds that even if you opt out of training, it trains on chats you give feedback on and chats "flagged for safety review" (Anthropic). For the training settings app by app, see which AI companies train on your data.

Does deleting a chat end the company's license?

Only Meta ties its license to deletion. Meta's Terms of Service say its content license "will end when your content is deleted from our systems," and that deletion can take up to 90 days, plus up to another 90 to clear backups (Meta). Mistral limits its own license to a purpose instead: it runs "for the duration necessary to achieve the intended purposes" (Mistral).

The others either say nothing on how long their rights last, or say they last forever. xAI's, Character.AI's, and Replika's licenses are perpetual and irrevocable. Google's "lasts for as long as your content is protected by intellectual property rights"; deleting stops Google "making that content publicly available," which covers things you posted, not a private chat (Google).

If a model has already learned from your chat, deleting the chat later doesn't pull it back out of the trained model, as we explain in what is machine unlearning. And because xAI's and Character.AI's licenses are "transferable," they can pass to whoever buys the company.

Can you use AI output commercially?

Mostly, with three kinds of limit.

1. Personal use only. Microsoft says "You can only use Copilot for your own personal use" (Microsoft), and Mistral's consumer terms let you use its products "solely in your personal capacity" (Mistral).
2. No training rival models. OpenAI bars using output "to develop models that compete with OpenAI," Anthropic bars using its services to "develop or train any artificial intelligence or machine learning algorithms or models" that compete, and Google bars "using AI-generated content from our services to develop machine learning models or related AI technology" (OpenAI, Anthropic, Google).
3. Labeling. OpenAI's terms forbid presenting output as "human-generated when it was not," and Mistral's say the same. xAI's terms say that "When using Output or SpaceXSI's name, logos, trademarks, or other brand elements, you are required to obtain our permission and attribute your generation of the Output to the Service" (xAI).

DeepSeek goes the other way. Its terms list "derivative product development, training other models (such as model distillation)" among the uses it allows (DeepSeek).

Can someone else get the same answer you did?

Yes, and the terms say so. OpenAI's say "other users may receive similar output from our Services" and that its assignment to you "does not extend to other users' output" (OpenAI). xAI, DeepSeek, Mistral, Microsoft, and Proton have similar clauses. Mistral's says your output "may be similar or identical" to another user's, and that theirs "are not considered your Output" (Mistral).

So owning an output doesn't stop anyone from getting the same words from the same model. If it matters that a piece is yours, the Copyright Office's report says a person can hold copyright in their own work that shows in an output, and in the "creative selection, coordination, or arrangement" or "creative modifications" of it (Copyright Office), so keep your drafts and edits.

How Morse keeps your chats yours

A license only reaches what the company holds. Every wide license above sits on a copy of your chats on the company's servers, and that copy is what can train a model, pass to a buyer, or answer a court order. Morse is built so it doesn't hold one.

• You own it, and the license is narrow. Morse's terms say "You own Your Content, as between you and us," and that you give Morse a license "only as needed to run the Service for you." They add: "We don't use Your Content to train AI models, and we don't sell it." There's no training switch to find.
• There's no server copy to license. Your chat history is stored only in your browser, encrypted, not on Morse's servers, and Morse doesn't keep messages or log message content. Deleting a chat or clearing your browser removes it for good, and Morse can't recover it either, so save a copy of anything you need.
• Each mode limits who else sees it. Use Anonymous for GPT, Claude, and Gemini without the provider knowing it's you: Morse's proxy removes your account, email, and IP address, though the provider may keep the message for a limited time to check for abuse, under its own policy. Use Private for open-weight models, where Morse has zero data retention agreements with its inference, search, and other providers, so they keep nothing. Use E2EE for anything no one should read: your browser checks the enclave's proof that it's genuine, then encrypts the message so Morse's servers relay text they can't read. This rests on the enclave hardware and attestation working as designed.

Morse doesn't show ads, and because it doesn't store chats, it usually has very little to give authorities.

Try a chat from the Morse home page without an account, or start free with no card. End-to-end encrypted chats come with Pro, at $19 a month.