
Is it true that Meta AI reads all your WhatsApp chats unless you turn on “Advanced Chat Privacy”?
検証日 · 更新
うわさ: A warning is being forwarded in WhatsApp groups: “starting today” (or “from Saturday”) Meta AI gets access to all chats, messages and phone numbers. To stop it, group admins supposedly have to turn on “Advanced Chat Privacy” in every chat and forward the warning.
1. 判定
Meta AI does not get access to all chats on any date. It only sees messages where someone tagged @Meta AI or that were forwarded to it, and chats themselves remain protected by end-to-end encryption (only participants can read them). “Advanced Chat Privacy” has existed since April 2025, but it protects you from other participants, not from a “hidden” AI: they can't export the chat, auto-save media or send messages from that chat to the AI.
High: WhatsApp's help center and blog state directly what the AI sees and what the setting does, and this matches the findings of about ten independent fact-checkers in different countries. Not the highest level, because nobody outside Meta has audited how processing works inside the company — we only know it from Meta itself.
2. 主張と事実
- «Starting today (from Saturday) Meta AI has access to all WhatsApp chats»虚偽
No. According to WhatsApp's help center, AI bots only read messages sent to them, and personal chats are end-to-end encrypted: “only chat participants can read and share them.” Meta confirmed to Chequeado, Verificat and Fact Crescendo that the AI only sees messages that tag @Meta AI or that a user chose to share with it. There was no “access switch-on” date — the chain has circulated since July 2025 with different dates.[1][3][4][5]
- «The AI can read messages and see participants' phone numbers and personal data»虚偽
No. Per WhatsApp's help center, as summarized by Fact Crescendo, Meta AI has no access to the full group chat or to participants' contacts. In a group, only a participant can start a conversation with the AI — not Meta or WhatsApp.[1][5][3]
- «To stop the AI reading a chat, the admin must turn on “Advanced Chat Privacy”»一部事実
Partly. The setting is real; WhatsApp introduced it on April 23, 2025. It stops other participants from exporting the chat, auto-saving media and “using messages for AI features” — so participants can't call Meta AI in that chat. But without it the AI still doesn't read the chat on its own: you only need it if you don't trust other participants. As Verificat puts it, it protects you “from other users, not from the AI.”[2][4][3]
- «The warning must be forwarded to every group»根拠なし
Forwarding protects nothing and only spreads an inaccurate text. Maldita, Chequeado and Zamora News note that this is a typical sign of a chain message.[10][3][9]
3. うわさの出どころ
- WhatsApp blog · en
WhatsApp introduces “Advanced Chat Privacy”: blocks exporting chats, auto-saving media and using messages for AI features. This real announcement later becomes the basis of the rumor.[2]
- Facebook, India · en
A post claims that without the setting chats are “vulnerable to Meta AI” and “all AI agents can access all chats.” The Quint rates it misleading.[6]
- WhatsApp, Spain · es
Spanish version: the AI will read all chats and numbers unless you turn on the setting. Verificat debunks it; Meta confirms personal messages are out of the AI's reach.[4]
- X, India · Vijay Shekhar Sharma, founder of Paytm · en
Post: “Today onwards WhatsApp is allowing AI to read chats,” turn on the setting. The same day WhatsApp replies publicly that Meta AI only reads what is shared with it. On August 20 Fact Crescendo rates the post false.[5]
- WhatsApp, Argentina · es
Version “Starting today AI is available in WhatsApp and has access to all chats,” with an “Enhanced Privacy” setting. Chequeado debunks it.[3]
- WhatsApp, Colombia · es
A “new viral chain” says the AI reads chats and groups. El Colombiano calls it false.[7]
- WhatsApp, Mexico · es
The chain returns: group admins are again told to turn on the setting. El Mañana quotes Meta calling the claim “completely false.”[8]
- WhatsApp, Spain · es
Version with a new date: “from Saturday AI will be available in WhatsApp and have access to all chats.” Zamora News debunks it.[9]
- ≈ Maldita.es, Spain · es
Maldita resurfaces its debunk: the AI only reads messages where it was mentioned or that were shared with it.[10]
- WWWhat's New, Spain · es
An explainer on what Meta AI sees in WhatsApp and how to limit it in the EU, citing Maldita's debunk. The topic is back in view.[11]
文面の変化
- April 2025: WhatsApp introduces a real setting that, among other things, stops participants sending chat messages to the AI
- → July 2025: “without this setting, all AIs have access to all chats”
- → August 2025: urgency added — “starting today”; the AI supposedly also sees numbers and personal data
- → November 2025: a “new chain” — “the AI reads chats and groups,” admins must turn on the setting
- → September 2026: a new date — “from Saturday”; a request to forward to every group
4. メディアの報道
- Chequeado (Argentina)否定
False. Meta: the AI only processes messages that tag it or that the user chose to share; personal chats are encrypted.[3]
- Verificat (Spain, Catalonia)否定
False. Meta's press office: personal messages are out of the AI's reach. The setting protects against other participants, not the AI. A lawyer notes that metadata (who you talk to and when) is a separate issue.[4]
- Fact Crescendo (India)否定
False. WhatsApp: Meta AI only reads what you choose to share with it.[5]
- The Quint (India)否定
Misleading: the setting is real, but without it the AI doesn't read chats.[6]
- El Colombiano (Colombia)否定
The new chain is false: Meta cannot read your messages.[7]
- El Mañana (Mexico), Zamora News (Spain)否定
- Maldita.es (Spain)否定
False: the AI can only read messages where it was mentioned or that were shared with it.[10]
- WWWhat's New (Spain)背景
What Meta AI really can do: anything sent to it (including your message, if another participant forwarded it) Meta may store and use. According to the article, you can't fully remove the AI from the app, but you can block it in a specific chat.[11]
6. 情報操作の兆候
- Tied to a real feature. “Advanced Chat Privacy” really has existed since April 2025 and mentions AI — this makes the text sound plausible.
- A precise, near date. “Starting today” (Chequeado, 2025) and “from Saturday” (Zamora News, 2026) — the date changes, yet access is never “switched on.”
- Call to act and forward. Admins are urged to turn on the setting right away and send the warning to every group (Zamora News).
- Unverifiable source. The text links to neither WhatsApp nor Meta; the company itself called the claim false (El Mañana, Fact Crescendo).
7. 過去の類似したうわさ
- “I do not give Facebook permission to use my photos and data” — a text people were told to copy onto their profile2012
結果: Has no effect: the terms of use are accepted at sign-up. The chain has kept returning for more than ten years; the last notable wave was in March 2024.[12]
過去のうわさとの共通点
- An urgent “forward to all groups” warning with no author and no link to the company
- Tied to a real feature or setting few people know about
- Promises protection through one simple action
新しい点
- Fear of AI: people don't understand what a built-in assistant actually sees
- The real setting does concern AI, but it protects against other participants, not the company
- The text lives for years and gets a new date each time (“starting today,” “from Saturday”)
9. どうすればよいか
- The AI doesn't read your chats on its own. But anything you or other participants send to Meta AI, or where it is tagged with @Meta AI, reaches Meta and may be used — don't put anything personal there.
- If you don't want participants sending chat messages to the AI or exporting the conversation, turn on “Advanced Chat Privacy”: tap the chat name and find it in the settings. It won't hurt.
- Got this warning? Don't forward it. Instead, briefly explain in the group that the setting protects against other participants, not against a “hidden” AI.
10. 確認できなかったこと
- The Colombiacheck page didn't open (the site blocked our tool), so we didn't see the date or text of their check; judging by the headline, it is the same check as Chequeado's.
- The Maldita.es debunk didn't fully load — we rely on its headline; the date in the page address (October 6, 2026) may mean an update rather than first publication.
- We did not ourselves find fresh forwards of the chain on November 9, 2026; the last confirmed wave was in September 2026 (Mexico, Spain).
- How message processing actually works inside Meta has not been checked from outside: it is known from the company and its help center.
- Metadata (who messages whom and when) is a separate topic; this check doesn't cover it.
11. 情報源
- WhatsApp Help Center — AI in chats and end-to-end encryption · 閲覧日 2026.11.09
- WhatsApp blog — Introducing Advanced Chat Privacy (Apr 23, 2025) · 閲覧日 2026.11.09
- Chequeado — No, Meta's AI cannot access WhatsApp chats automatically (Aug 22, 2025) · 閲覧日 2026.11.09
- Verificat — Meta's AI cannot access all your WhatsApp chats automatically (Jul 25, 2025) · 閲覧日 2026.11.09
- Fact Crescendo — Will AI Access Your WhatsApp Chats Without “Advanced Chat Privacy” Enabled? (Aug 20, 2025) · 閲覧日 2026.11.09
- The Quint — No, Meta AI Can’t Read All Individual and Group WhatsApp Chats (Jul 16, 2025) · 閲覧日 2026.11.09
- El Colombiano — New viral chain about “WhatsApp AI that reads chats and groups” is false (Nov 5, 2025) · 閲覧日 2026.11.09
- El Mañana — WhatsApp: can Meta AI read your chats if you don't turn on this feature? (Sep 11, 2026) · 閲覧日 2026.11.09
- Zamora News — Hoax check: AI will be able to read all your WhatsApp chats from Saturday (Sep 20, 2026) · 閲覧日 2026.11.09
- Maldita.es — It's false that Meta's AI can access “all” your WhatsApp messages if you don't turn on advanced privacy · 閲覧日 2026.11.09
- WWWhat's New — Meta AI in WhatsApp: what it can see in your chats and whether it can be removed (Nov 9, 2026) · 閲覧日 2026.11.09
- Malwarebytes — New Facebook photo rule hoax spreads (Mar 13, 2024) · 閲覧日 2026.11.09
5. SNS
The rumor lives in forwarded group messages, so its reach can't be measured. In 2025–2026 it was debunked in India, Spain, Argentina, Colombia, Mexico and other countries.[5][3][7][8][9]
A public post saying “WhatsApp is allowing AI to read chats” and urging people to turn on the setting; WhatsApp replied the same day that this is not so.[5]
2025-08-19