Are AI chats private? ChatGPT, Claude, Gemini, Grok, and Meta AI compared

The Morse teamPublished October 15, 20268 min read

No, not in the way most people mean. AI chats are private from other users unless you share them, but ChatGPT, Claude, Gemini, Grok, and Meta AI all keep your chats on their servers, let some staff or contractors read some of them, and hand them over when the law compels it.

They differ on whether your chats train the next model by default, how long they're kept after you delete them, whether they pick the ads you see, and what happens when a court gets involved. We read each company's own policy and help pages on October 15, 2026 and put the answers side by side below.

How ChatGPT, Claude, Gemini, Grok, and Meta AI compare

This covers personal accounts (free and paid consumer plans), not business or API plans. Every cell links to the company's own page, read October 15, 2026. Policies change often, so check the date.

No consumer AI chat service that stores your conversations can refuse a valid court order for them, so the last column reads the same for every company.

Which AI companies train on your data?

All five, by default, on personal accounts. The difference is how you get out.

• ChatGPT: turn off "Improve the model for everyone" in Settings, Data controls. OpenAI notes that if you rate a reply with a thumbs up or down, "the entire conversation associated with that feedback may be used to train our models," even after you opt out (OpenAI).
• Claude: Anthropic asked every consumer user to choose in 2025, and keeps chats for five years instead of 30 days if you allow training (Anthropic, August 28, 2025). Even if you opt out, chats "flagged for safety review" and feedback you send can still be used (Anthropic).
• Gemini: turning off Keep Activity stops training, but it also turns off your chat history and some connected apps (Google).
• Grok: turn off "Improve the model" in Data Controls. Grok inside X follows X's own policy and settings instead (Grok FAQ).
• Meta AI: Reuters reported in October 2025 that people using Meta AI "will not have an option to opt out" of their chats being used for ads (Reuters via The Globe and Mail). Meta's own pages don't describe a training opt-out for US users that we could find.

Four of the five offer a temporary or incognito mode that's kept out of training: ChatGPT's Temporary Chat, Claude's Incognito chats, Gemini's temporary chats, and Grok's Private Chat. Business and enterprise plans are excluded by default at OpenAI, Anthropic (Anthropic), and Grok's maker.

Do humans read your AI chats?

Yes, some of them, at every company in the table. It's how chatbots are graded and improved, and how abuse is caught.

Google is the most specific. Its reviewers "assess if Gemini Apps' responses were low-quality, inaccurate, or harmful," and reviewed chats are disconnected from your account and kept for up to three years (Google). Anthropic says its employees "cannot access your conversations" by default, outside feedback you send and Trust and Safety reviews (Anthropic).

In September 2026, 404 Media reported, from leaked documents, that OpenAI contractors read real ChatGPT conversations to rate replies, under an effort called Project Lily. Reviewers don't see usernames, and according to 404 Media, OpenAI "acknowledged sensitive details can still get through" (404 Media, September 14, 2026). A proposed class action over the program was filed in federal court in Northern California; its claims are unproven (Decrypt). Anthropic told 404 Media it also uses human review to improve its models.

How long do AI companies keep your chats after you delete them?

Deleting a chat removes it from your sidebar at once. The copy on the company's servers goes later:

• ChatGPT, Claude, and Grok: within 30 days, with exceptions for legal and safety reasons.
• Gemini: deleted from your activity, but any chat a human already reviewed is kept for up to three years, because it's no longer connected to your account (Google).
• Meta AI: Meta's AI terms say "the AIs will sometimes retain and use that information" and ask you not to share "information about sensitive topics" (Meta).

Anthropic says turning training off stops future use, but your data "will still be included in model training runs that are already in progress, or in models that have been trained" (Anthropic).

Are Meta AI chats private?

From other people, yes, unless you share them. In 2025, Meta's AI app published shared chats to a Discover feed, and Fortune reported that many people didn't realize what they'd posted there (Fortune). From Meta, no: since December 16, 2025, Meta uses "interactions with AI at Meta to personalize the content and ads you see," except for sensitive topics like health, religion, sexual orientation, and politics (Meta, October 1, 2025).

Meta is ahead in one place. Its AI terms say AI chats that use its Private Processing technology are ones "Meta cannot read or access" (Meta).

Can your AI chats be used against you in court?

Yes. In United States v. Heppner, Judge Jed Rakoff ruled on February 13, 2026 that a defendant's conversations with Claude were not protected by attorney-client privilege or work product (Orrick). In the New York Times copyright case, a judge affirmed on January 5, 2026 that OpenAI must produce 20 million de-identified ChatGPT logs, finding users' privacy protected enough by de-identification and a protective order (Jones Walker).

Many guides still say OpenAI keeps deleted chats "indefinitely," but that duty ended on September 26, 2025, though OpenAI still holds some data from April to September 2025 (OpenAI). The Heppner chats also didn't come from Anthropic. The FBI seized about 31 documents recording them during a search of the defendant's home, so what's on your own devices counts too. We cover this in more depth in can AI chats be subpoenaed. This isn't legal advice.

How to keep your AI chats more private

1. Turn off training in each app's data settings, and don't rate replies you'd rather keep out of training.
2. Use temporary modes for sensitive questions, knowing they're still kept for 72 hours to 30 days.
3. Don't share links to chats with anything personal in them.
4. Leave out names, account numbers, and health details where the question works without them.
5. Pick a tool whose design limits what's kept, not only its settings, like Morse below. Our guide to using AI privately covers more ways to do it.

How Morse keeps your AI chats private

Every risk in the table above starts with your chats sitting on a company's servers, tied to your account. Morse is built so they don't. It's a web app with over 30 models, open and closed, in one place, and your chat history is stored only in your browser, encrypted, not on our servers. We don't train AI on your content, sell your data, or show ads (privacy policy).

Each chat has a privacy mode, so you choose how private each conversation is:

• Anonymous, to use GPT, Claude, and Gemini without the provider knowing it's you. Our proxy removes your account, email, and IP address before the message goes out, and sends it under Morse's own account. The provider sees the question, with your identity stripped.
• Private, for anything you'd rather no host keep. It's the default for a new chat. Open-weight models like DeepSeek, Kimi, and GLM, plus Grok, run on hosts that have agreed to zero data retention. Where a host confirms that on each reply, Morse checks it and refuses a reply that doesn't.
• E2EE, for anything no one should read. Your browser checks the secure enclave's proof that it's genuine before anything leaves your device, then encrypts each message so only that verified enclave can open it. Our servers relay text they can't read, and the reply is opened only in your browser. Every E2EE reply has a lock you can open to see each check that passed. Read how end-to-end encrypted AI chat works.

There's no training on your chats to opt out of, no ads picked from them, and no chat history on our servers for anyone to read. Because we don't store chats, we usually have very little to give authorities when a court asks. Your history is locked when you sign out and takes two keys to open: one that never leaves your device, and one our server gives you when you sign in. Clearing your browser or deleting your account deletes your chats for good, with no server copy left behind. For a question you don't want saved anywhere, a Temporary chat is never saved, not even on your device.

Try a chat from the Morse home page without an account, or start free. End-to-end encrypted chats come with Pro, at $19 a month.