Is Meta AI private? Ads, training, Incognito Chat, and deleting Meta's copy

The Morse teamPublished October 19, 20269 min read

As of October 2026, Meta AI is private from other people unless you share your chats, but not from Meta, which keeps those chats, can train its AI on them, may have people review them, and since December 16, 2025 uses them to pick the ads and posts you see on Facebook and Instagram. The one exception is Incognito Chat, launched on May 13, 2026, which Meta says runs where even Meta can't read it.

We read Meta's AI terms, its privacy policy, six Privacy Center pages on generative AI, five Meta and WhatsApp help pages, and four Meta newsroom posts on October 19, 2026. Every fact below links to the page it came from. Meta deserves credit in two places. Its Incognito Chat runs in a secure enclave and lets you download a report of what was processed, and its help pages give exact menu paths for deleting things.

Meta AI privacy at a glance

These are personal accounts in the US, where most of the rules below apply. Regions differ, and we cover that further down.

Does Meta AI train on your messages?

Yes, on what you send to Meta AI: its Privacy Center says "your interactions with AI features can be used to train AI models," including "messages to AI chats" (Meta), and we found no training switch for US users on any Meta page we read. We compare Meta's training rules with other companies' in which AI companies train on your data.

Are Meta AI chats used for ads?

Yes, since December 16, 2025. Meta announced on October 1, 2025 that "we will soon use your interactions with AI at Meta to personalize the content and ads you see, including things like posts and reels," with notices going out from October 7 (Meta). Its own example is a chat about hiking that leads to "ads for hiking boots."

The same post sets three limits.

• Sensitive topics are left out of ads. Chats about "religious views, sexual orientation, political views, health, racial or ethnic origin, philosophical beliefs, or trade union membership" aren't used to show you ads.
• Accounts Center decides how far it reaches. If your WhatsApp account isn't in the same Accounts Center as Facebook or Instagram, "interactions with Meta AI on WhatsApp will not be used to personalize experiences across different accounts."
• Voice counts too. The change covers "a voice chat or a text exchange."

You can't switch it off. Reuters reported that users "will not have an option to opt out" (Reuters via The Globe and Mail, October 1, 2025). Meta points to Ads Preferences and feed controls, which adjust the ads you see but don't stop your chats feeding them. That exclusion applies to ad picks only, and Meta's AI terms ask you not to share "information about sensitive topics" at all (Meta).

Can other people see your Meta AI chats? The Discover feed, links, and vibes

Only what you post or share. When Meta launched the Meta AI app on April 29, 2025, it included a Discover feed "to share and explore how others are using AI," and Meta said "nothing is shared to your feed unless you choose to post it" (Meta). Many people posted personal chats there without realizing it, as we note in are AI chats private.

As of October 2026, Meta's help pages describe two ways a chat leaves your account:

1. Share links. In a chat, tap More, then Share a link. Anyone with the link sees the chat up to the moment you made it, and "the link doesn't expire and people can share the link" (Meta). To kill it, open the chat, tap Menu at the top right, then Delete link. Deleting the chat also breaks the link.
2. Public vibes. AI videos you post show up in Feeds and on your public profile. To remove them all, tap Menu, then Settings, then Data & privacy, Manage your information, and Remove all public vibes (Meta).

We didn't find a current Meta help page that describes posting text chats to the Discover feed, so check your app's History for anything marked public if you used it in 2025. Incognito chats "can't be shared" at all (Meta).

Is Meta AI on WhatsApp private?

Your personal WhatsApp chats are; your chats with Meta AI follow Meta's rules. WhatsApp says "only messages that mention @Meta AI, or that people choose to share with Meta AI, can be read by Meta," and Meta receives "your prompts, AI messages, and feedback" to answer you and "to improve its AI quality" (WhatsApp). The same page says Meta trains its models on data "including the messages and prompts you send to Meta AI."

Some WhatsApp AI features use Private Processing instead. Message summaries, writing help, and Incognito chats run in a Trusted Execution Environment "where no one, not even Meta or WhatsApp, can read or access your messages," and the messages "aren't stored" afterward (WhatsApp). You can download a Private Processing report that's made on your device. To keep others from using these features in a chat, turn on Restricted chat: a group admin can, and in a one-to-one chat either person can. As of October 2026, Private Processing supports English, Indonesian, Portuguese, and Spanish.

Is Meta's Incognito Chat really private?

It's the most private way to use Meta AI. Meta launched Incognito Chat on WhatsApp and the Meta AI app on May 13, 2026, saying your messages "are processed in a secure environment that even Meta cannot access" and "your conversations are not saved" (Meta). In the Meta AI app, tap Menu, then Incognito chat at the top, then Ask Meta AI privately… (Meta).

It has two catches. Web search is on by default, and Meta AI turns your prompt into a search query sent to search engines "without connecting the search to your account," which you can turn off (Meta). And the chat is gone when you exit, so there's no history to come back to. Meta said Incognito Chat would roll out "over the coming months," so it may not be in your app yet.

How do you delete Meta AI chats for good?

It depends on the app, and deleting what you see doesn't always delete Meta's copy. Meta's AI terms, effective May 13, 2026, warn that "deleting individual messages, entire threads, or even your account may not delete our copy of your personal information" (Meta).

• Meta AI app: tap Menu, then Settings, Data & privacy, Manage your information, and Delete all chats and media. Chats leave your History and Media, but public vibes stay up until you remove them separately (Meta).
• WhatsApp, Messenger, and Instagram: type /reset-ai in a chat with an AI to delete its copy of that chat, or /reset-all-ais to reset every AI chat on that app. Meta says on WhatsApp "if you delete a chat with AI in the WhatsApp app, the information will only be deleted on WhatsApp. To delete it from Meta, you'll need to use the reset command" (Meta).
• Voice chats: your recording stays in history only "temporarily," but the text transcript remains until you delete it (Meta).

Run the reset command before you delete the chat on your phone, since the command is typed into that chat. On October 19, 2026, the top Google results for this query were mostly about the Discover feed, and none of them led with the reset command.

Does it depend on where you live?

Yes. Meta's rules for AI chats differ by region, and the AI terms point people in the UK, Brazil, and the European Region to their own versions (Meta).

Meta said in October 2025 it hoped to bring the ad change "everywhere soon," so the exclusions may not last.

Who else can get your Meta AI chats?

People at Meta, its vendors, and courts. Meta's terms say it uses AI content "through automated or manual (i.e. human) review and through third-party vendors in some instances" (Meta). When Meta AI can't answer, it shares "the messages you send to AI and general information, like your region, with select partners" (Meta). And the privacy policy says Meta shares information "in response to legal requests, like search warrants, court orders, production orders or subpoenas" from "civil litigants, law enforcement and other government authorities" (Meta). We cover how that has played out with AI chats in can AI chats be subpoenaed. This isn't legal advice.

How Morse keeps your questions out of an ad profile

Everything above comes back to one design. Your Meta AI chats sit on Meta's servers, tied to the same identity as your Facebook, Instagram, and WhatsApp accounts, and that identity is what Meta sells ads against. Morse is built without either piece: your chat history lives in your browser, and Morse doesn't show ads, train AI on your content, or sell your data (privacy policy). Morse is a web app with over 30 models, open and closed, and each chat gets one of three privacy modes.

• Anonymous, for questions you'd never want tied to your profile. Ask GPT, Claude, or Gemini, and our proxy removes your account, email, and IP address before the message goes out under Morse's own account. The provider sees the message without knowing it's you, and may keep it for a limited time to check for abuse, under its own policy. Leave out names you don't want anyone to see, since the text itself still reaches the provider.
• Private, for chats you'd rather no company keep. It's the mode a new chat starts in. Open-weight models like DeepSeek, Kimi, and GLM, plus Grok, run here, and Morse has zero data retention agreements with its inference, search, and other providers. Where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
• E2EE, for the conversations Incognito Chat was built for. Meta's Incognito Chat and Morse's E2EE both use secure enclaves. On Morse, your own browser checks the enclave's hardware attestation before anything leaves your device, then encrypts the message so only that enclave can open it, and our servers relay text they can't read. The enclaves are run by third-party providers, not by Morse, and the protection rests on the enclave hardware and attestation working as designed. See how end-to-end encrypted AI chat works.

Your history is stored in your browser, encrypted, and locked when you sign out. Opening it takes two keys, one that never leaves your device and one our server gives you when you sign in. There's no ad system for your chats to feed, no advertising trackers or third-party analytics, and no reset command to remember. Because we don't store chats, we usually have very little to give authorities. Clearing your browser or deleting your account removes your chats for good, and Morse can't recover them either, so keep a copy of anything you need. For a question you don't want kept on your device, a Temporary chat is held in memory only.

So is Meta AI private enough for you?

• Ads: in Meta AI, keep health, money, and relationship questions out of normal chats, and keep WhatsApp out of your Accounts Center. On Morse, there are no ads to pick.
• Training: in Meta AI, use Incognito Chat where you have it. Morse doesn't train AI on your chats.
• Public posts: check your share links and public vibes in the Meta AI app, and copy text instead of sharing links.
• Meta's copy: type /reset-ai before you delete a chat on WhatsApp, Messenger, or Instagram. On Morse, your device holds the only copy.

For more habits that work in any app, read how to use AI privately.

Send a first message from the Morse home page with no account, or start free. End-to-end encrypted chats come with Pro, at $19 a month.