Can Google’s human reviewers see my Gmail and Drive through Gemini?

Yes, if you’ve connected Google Workspace to Gemini through Personal Intelligence on a personal account. Google says “a subset of Connected Apps data processed by Gemini is reviewed by human reviewers,” including excerpts from your emails and files, and that “the summary may just be the email or file” when it’s brief or particularly relevant. Google says reviewers may also look at some of your inbox if you send feedback or it needs to address abuse or harm.

Did Google change its policy?

Yes, in January 2026, when Personal Intelligence launched. Until then, Gemini’s privacy hub said content from Google services was “not allowed to be accessed or reviewed by human reviewers, except for safety and security purposes.” On October 16, 2026, Google added a note to the two 2023 posts that made the promise. A GitHub write-up brought it to wide attention on November 9.

How do I disconnect Gmail and Drive from Gemini?

In Gemini, open Settings, then Personal Intelligence, then Connected Apps, and disconnect Google Workspace. Google Search’s AI Mode connects separately. To cut Workspace off from both, turn off Smart features in other Google products in Gmail’s settings.

Does disconnecting delete what reviewers have already seen?

No. Google says disconnecting “doesn’t delete any data from Gemini Apps Activity,” and chats that reviewers have seen are kept for up to three years even if you delete your activity. Deleting the chats that used your email keeps any not yet reviewed out of future product improvement.

Does this apply to my work or school account?

No. Personal Intelligence “isn’t available when you sign in to a work, school, or supervised Google Account,” and Google’s Workspace privacy hub says chats and uploaded files in the Gemini app “won’t be reviewed by human reviewers or otherwise used to train generative AI models without your permission.”

Do ChatGPT and Claude train on the Gmail I connect to them?

Their policies differ. OpenAI says it “may use information accessed from apps to train our models” on Free, Plus, Go and Pro if Improve the model for everyone is on. Anthropic says the chats it may train on don’t include “raw content from connectors (e.g. Google Drive),” unless it’s copied into the conversation.

Inter-office memo

To: Everyone who has connected Gmail, Drive or Photos to Gemini
From: Mo, Chief Executive
Date: November 11, 2026, 9:20 AM CST
Re: Human review of Gemini’s Connected Apps

Google’s reviewers can see Gmail and Drive data you connect to Gemini. How to stop it

If you connect Google Workspace to Gemini through Personal Intelligence on a personal account, Google says “a subset” of that data goes to human reviewers, including excerpts and sometimes whole emails or files. Its help page has said so since January 2026, and on October 16 Google amended a 2023 promise that reviewers wouldn’t see Gmail, Docs or Drive. Disconnecting stops it for future chats.

What Google says reviewers see

Google’s Connected Apps page answers the question directly. Under “Will human reviewers have access to my data?”, it says: “When connected to Gemini, a subset of Connected Apps data processed by Gemini is reviewed by human reviewers (including trained service providers) to debug our services, help keep them safe, and improve them.”

It lists what that subset includes:

Your email and files. “Interactions, summaries, excerpts and inferences from your relevant emails and files.” And: “In some cases, for example, where the email or file is brief or particularly relevant to respond to the prompt, the summary may just be the email or file.”
Your photos, indirectly. “Summaries, inferences, and generated media based on your Google Photos library.” Google adds that “except for cases of abuse and harm or when you submit feedback, humans don’t view media from your Google Photos library.”
Your Wallet. “Interactions, summaries, excerpts, and inferences from your Google Wallet and payment services information.” Wallet, which Google says can include “financial account info linked with services such as Plaid,” wasn’t on this list in an archived copy from September 17.
The chat itself. Your prompts, Gemini’s responses and “information about how our models generated those responses.”

Before outside reviewers see it, Google says it takes “steps to protect your privacy, such as disconnecting interactions from your Google Account and reducing personal information in this data, unless it’s to address abuse or harm.” One sentence goes further: “If you submitted feedback, or we need to address abuse or harm, human reviewers may also review some of your Gmail inbox or Photos library media.” The privacy hub says reviewed chats “are not deleted when you delete your activity” and are “retained for up to three years.”

Google’s own advice is on the same page: “Please don’t connect your apps to Gemini Apps if they contain confidential information that you wouldn’t want a reviewer to see.”

What changed, and when

Little of this is new this week. Google’s pages, and archived copies of them, show when each piece arrived:

Date	Page	What it said
September 19, 2023	Google’s Bard extensions post	“Your content from Gmail, Docs and Drive is not seen by human reviewers, used by Bard to show you ads or used to train the Bard model”
October 12, 2023	Google’s post on generative AI guardrails	The same promise, nearly word for word
January 13, 2026	Gemini’s privacy hub, the day before the launch	Content from Google services is “Not allowed to be accessed or reviewed by human reviewers, except for safety and security purposes”
January 14, 2026	Personal Intelligence launch	A beta for US Google AI Pro and Ultra subscribers. “Connecting your apps is off by default”
January 15, 2026	The Connected Apps page	“A subset of Connected Apps data processed by Gemini is reviewed by human reviewers”
March 17, 2026	Expansion post	Personal Intelligence reaches AI Mode in Search, and starts reaching free Gemini users in the US
October 16, 2026	Both 2023 posts	An update note: if users turn on newer features, Workspace data “can be used to train generative AI models” and “can also be subject to human review in limited circumstances”
November 9, 2026	GitHub write-up	Posted to Hacker News, where it drew 723 points

Google’s January and March posts describe training (“Gemini doesn’t train directly on your Gmail inbox or Google Photos library”) but don’t mention human review. The 2023 promise stayed on Google’s blog, unqualified, until the October 16 note, which links to Search’s version of the policy. An archived copy of the Bard post from October 7 has no note.

The write-up’s author describes tests on a personal Gmail account in which Gemini pulled emails into ordinary chats without being asked to open Gmail, sometimes repeating account and confirmation numbers, and at times “the entire text dump of the email, including its headers.” Those are one person’s tests. Google doesn’t describe it that way, but its setup page says that once you’ve connected apps, “if you start a new chat, Personal Intelligence is on by default.”

Who it affects

It applies to you only if all of these are true, per Google’s setup page:

You use a personal Google account and you’re 18 or over. Personal Intelligence “isn’t available when you sign in to a work, school, or supervised Google Account.”
You opted in and connected an app. Google reset earlier connections when Personal Intelligence launched: “If you’ve previously connected the Google Workspace app or the Google Photos app, your settings have been reset.”
Keep Activity is on. Personal Intelligence “isn’t available when this setting is off.”
You’re outside the European Economic Area, Nigeria, Switzerland and the UK, where the feature isn’t offered.

Work and school accounts are covered by their organization’s terms. Google’s Workspace privacy hub says “Your chats and uploaded files in Gemini app won’t be reviewed by human reviewers or otherwise used to train generative AI models without your permission.”

Search is a second door. AI Mode and Ask Maps can connect to Gmail, Calendar and Photos, and Google’s Search help uses the same words: “a subset of that data processed by AI Mode in Search and Ask Maps in Google Maps is reviewed by human reviewers.”

How to check and disconnect

Disconnect in Gemini. On the web, open the menu, then Settings, Personal Intelligence and Connected Apps, and choose More options, then Disconnect next to Google Workspace. In the mobile app, it’s the menu, your profile picture, Personal Intelligence, then Connected Apps. Do the same for Photos, Wallet, Search services, YouTube and Contacts if you don’t want them used.
Disconnect in Search too. It’s a separate setting: in your Google Account, open Search Services Personalization, then Additional Personal Intelligence, then Connected Content Apps. Google notes it’s “currently updating settings for Google Search services,” so the names may differ for you.
Or use one switch for Workspace. In Gmail, open Settings, See all settings, then under General, Manage Workspace smart feature settings, and turn off Smart features in other Google products. Google says that turns off Personal Intelligence in the Gemini app and in Search, along with restaurant reservations in Maps and suggested tickets and passes in Wallet. Its Gemini page puts it more narrowly: with the setting off, “Gemini Apps can’t connect to Google Workspace.” Photos, Wallet and YouTube are still set in Gemini.
Delete the chats that used your mail. Google says you need both steps: “If you only disconnect the app, Gemini might still use the info if it’s in a past chat.”
Don’t send feedback on a chat that used your mail, since feedback is one of the cases in which reviewers “may also review some of your Gmail inbox.”
For one sensitive chat, turn it off there. Under Tools in the text box, turn Personal Intelligence off, or use a temporary chat.

How ChatGPT and Claude compare

On personal plans, the three labs’ pages say this about content from your connected apps:

Service	Does it train on content from your connected apps?
Gemini	On “summaries, excerpts, generated media, and inferences,” which may be “just the email or file,” and a subset is human reviewed
ChatGPT Free, Plus, Go, Pro	“OpenAI may use information accessed from apps to train our models if your ‘Improve the model for everyone’ setting is on”
Claude	The chat data it may train on “does not include raw content from connectors (e.g. Google Drive),” though it may “if it’s directly copied into your conversation”

Gemini’s is the only one of the three pages we read that describes people reviewing connected-app content.

What Google hasn’t said

How big “a subset” is. Its pages give no share or number.
Anything about the write-up. We found no statement from Google about it, and no one in the Hacker News thread says they speak for Google.
Why the 2023 promise stood until October. The note doesn’t say when it stopped applying.

Enjoy doesn’t run the Gemini app, so these switches are yours to set at Google (what AI providers see). For work you’d rather keep on your computer, OpenCode with a local model runs in Enjoy, and the model runs on your machine.

We’ll update this memo if Google responds or changes its pages.

7.Questions.

Can Google’s human reviewers see my Gmail and Drive through Gemini?: Yes, if you’ve connected Google Workspace to Gemini through Personal Intelligence on a personal account. Google says “a subset of Connected Apps data processed by Gemini is reviewed by human reviewers,” including excerpts from your emails and files, and that “the summary may just be the email or file” when it’s brief or particularly relevant. Google says reviewers may also look at some of your inbox if you send feedback or it needs to address abuse or harm.
Did Google change its policy?: Yes, in January 2026, when Personal Intelligence launched. Until then, Gemini’s privacy hub said content from Google services was “not allowed to be accessed or reviewed by human reviewers, except for safety and security purposes.” On October 16, 2026, Google added a note to the two 2023 posts that made the promise. A GitHub write-up brought it to wide attention on November 9.
How do I disconnect Gmail and Drive from Gemini?: In Gemini, open Settings, then Personal Intelligence, then Connected Apps, and disconnect Google Workspace. Google Search’s AI Mode connects separately. To cut Workspace off from both, turn off Smart features in other Google products in Gmail’s settings.
Does disconnecting delete what reviewers have already seen?: No. Google says disconnecting “doesn’t delete any data from Gemini Apps Activity,” and chats that reviewers have seen are kept for up to three years even if you delete your activity. Deleting the chats that used your email keeps any not yet reviewed out of future product improvement.
Does this apply to my work or school account?: No. Personal Intelligence “isn’t available when you sign in to a work, school, or supervised Google Account,” and Google’s Workspace privacy hub says chats and uploaded files in the Gemini app “won’t be reviewed by human reviewers or otherwise used to train generative AI models without your permission.”
Do ChatGPT and Claude train on the Gmail I connect to them?: Their policies differ. OpenAI says it “may use information accessed from apps to train our models” on Free, Plus, Go and Pro if Improve the model for everyone is on. Anthropic says the chats it may train on don’t include “raw content from connectors (e.g. Google Drive),” unless it’s copied into the conversation.