Google can train on your Gemini chats even if you pay. How to stop it, and what you lose

What set this off

On November 4, 2026, a post on Hacker News titled “Tell HN: Gemini trains on chat history, even for subscribers (personal accounts)” passed 300 points within hours. Its author wrote: “I learned this too late, thinking that having a paid subscription = not being turned into the product. Apparently paying is not enough.”

Nothing changed this week. The post quotes Google’s own help pages, and BGR walked through the same switch in August 2025, under its old name, Gemini Apps Activity. What’s new is the attention.

The post reads those pages correctly. Google’s Gemini Apps privacy hub has no separate rule for subscribers. Its only mention of a paid plan is in the list of what Google collects: “if you have a paid subscription to a Google AI plan, subscription related information.” Google’s plans page does promise “Your data—including prompts, outputs, and training—isn’t used to train Google models,” but that line belongs to Gemini Enterprise, its business editions.

What Google says it does

This covers the Gemini app on a personal Google account, on the free tier or on Google AI Plus, Pro or Ultra.

• It’s on by default. “If you’re 18 or over, Keep Activity is on by default,” says Google’s activity page.
• It includes training. With it on, “Google uses your activity to provide, develop, and improve its services (including training generative AI models).” That means “your prompts,” “the responses and suggestions that Gemini Apps give you,” and how the models produced them.
• Uploads count. With it on, “your screenshots and uploads of things like your media and files could be used to improve Google services with the help of human reviewers,” Google says.
• Connected apps count, indirectly. “Gemini does not train generative AI models directly on your Gmail inbox, Drive, Contacts, Calendar or other Workspace apps.” It trains on “summaries, excerpts, generated media, and inferences” made while answering you, and “the summary may just be the email or file” when that item is brief or particularly relevant. Google says it takes steps to reduce personal information first.
• People read some of it. “A subset of chats are reviewed by human reviewers (including Google’s trained service providers).” Reviewed data “is disconnected from your account and saved for 3 years,” and deleting your activity doesn’t remove it. Google’s advice: “Please don’t enter confidential information that you wouldn’t want a reviewer to see or Google to use to improve our services, including machine-learning technologies.”
• Voice is the exception. “Your audio and Gemini Live videos and screenshares aren’t used to improve Google services by default.” That has its own setting, off unless you turn it on.
• Ads aren’t part of it, for now. “Your Gemini Apps chats are not being used to show you ads. If this changes, we will clearly communicate it to you.”

How to turn it off, and what you lose

1. Open the setting. At gemini.google.com, select Settings & help, then Activity, or go straight to Gemini Apps Activity.
2. Turn it off. Near the top, select On, then Turn off, or Turn off and delete activity.
3. Decide about older chats. The switch covers “future chats.” Past ones stay in play until you delete them: “You can delete your Gemini Apps activity to prevent it from being used for future product improvement, unless it has already been reviewed by trained reviewers.” If you want a copy first, Google points to Takeout for exports.

Here’s what that costs you:

• Your history. “Your future chats won’t appear in your Activity.” Google says it keeps them with your account for up to 72 hours, to respond and to protect its services.
• Most Connected Apps. With it off, “only the following apps are available”: Device assistance, Gemini Notebook, Messages, Phone and WhatsApp.
• Chats with Gems. Your Gems stay, but “any chats you create with Gems will not be saved.”

And here’s what it doesn’t stop:

• Safety review. “Even if your Keep Activity setting is off or you use temporary chats, Google still uses your chats to respond to you and help protect Google, our users, and the public, including with help from human reviewers.”
• Anonymized data. “Your Gemini settings don’t control processing of your chats to create anonymized data to improve Google services.”
• Feedback. If you send feedback with it off, Google collects it along with “the last 24 hours of your chats” and any content in them, and keeps reviewed feedback for up to 3 years.

Your other options

1. Use temporary chats for the sensitive ones. “Temporary chats are not used to train Google’s AI models,” so you can leave Keep Activity on and history intact for everything else. They can’t use Gems, Connected Apps or personalization, you can’t send feedback from them, and they aren’t available on work or school accounts or in Gemini in Chrome or Messages. Google still keeps them for 72 hours.
2. Use a work account. On Workspace, “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,” Google says. The catch is that your administrator controls retention and activity settings.
3. Use the Gemini API on a paid project. With billing on, “Google doesn’t use your prompts… or responses to improve our products,” per its API terms, though it logs them for a limited time to detect abuse. The free quota and free AI Studio are the opposite: Google uses that content, and “human reviewers may read, annotate, and process” it, except in the European Economic Area, Switzerland and the UK, where the paid terms apply to all of it. One recent change to know: since October 26, new paid projects store Generate Content API requests by default, for up to 55 days. Those logs are “private to your Google Cloud project” and aren’t trained on unless you share them. To opt out, set store=false or turn logging off in AI Studio.
4. Use a local model. Nothing goes to Google at all, at the cost of the frontier models. See our guide to local coding agents.

How ChatGPT, Claude and Antigravity compare

Here are the personal plans, as of November 4, 2026:

Service	Trained on?	Where to turn it off	Keep your history?
Gemini, any personal plan	“On by default” from age 18	Settings & help > Activity	No. New chats last 72 hours
ChatGPT Free, Plus or Pro	“Data sharing is enabled by default”	Settings > Data controls	Yes: “turn off Improve the model for everyone and keep your chat history”
Claude Free, Pro, Max or Ultra	“When this setting is on”	Privacy settings	Yes. Anthropic’s pages treat it apart from chat history
Antigravity, any personal plan	“We use Interactions to evaluate, develop, and improve” Google’s products and machine learning	Settings, under Account: Enable Telemetry	Google’s docs don’t connect the two

Of the three chat apps, Gemini is the one where opting out of training means giving up your history. All three leave the choice to you, and none exempts you for paying. Our memo on Codex’s opt-out covers OpenAI’s side.

The switches also reach back differently. Anthropic says that with its setting off “we will not use your previous or new chats or coding sessions for future model training,” though conversations flagged for safety review may still be used. Google’s covers only future chats, so past ones have to be deleted.

If you code with Antigravity

Your Google AI plan pays for both products, but the switches are separate. Gemini’s privacy hub never mentions Antigravity, and Antigravity’s terms, settings page and FAQ never mention Keep Activity.

Antigravity’s terms say “We use Interactions to evaluate, develop, and improve Google and Alphabet research, products, services and machine learning technologies,” and that “Google employees and contractors may access, view, review and use Interactions.” Interactions are “your user data, interaction data pertaining to your usage of the Service, related metadata connected to the Service, and any feedback you provide.” The opt-out, in the same paragraph: “navigate to settings to change your preference on how such data is used.”

The setting is Enable Telemetry. “When toggled on, Antigravity collects interactions for use in evaluating, developing, and improving Antigravity and models that support Antigravity.” In the CLI, the choice is “saved with your account,” and it’s “always off for business accounts.” The docs don’t say what the default is on a personal account, so go and look.

Turn it off, then treat the result with some care. An account labeled Google wrote on Google’s forum in March that this “stops the collection of interactions data and diagnostic information used to improve the service.” Users have kept asking whether that also rules out training and human review, and a paying user asked again on October 22: “A clear yes/no, please.” That thread has no answer yet. The terms add: “Even when your Telemetry is toggled off, Google still uses your data to provide the service, process any feedback and help keep Antigravity safe.”

If you need that promise in a contract, access through Gemini Enterprise or a Workspace subscription falls under the terms your administrator accepted, not these.

What Google hasn’t said

• Whether it will respond. We found no statement from Google about the post, and no one in the thread says they speak for Google.
• How much gets reviewed. Its pages say “a subset” and give no number.
• Why training and history share a switch. Google explains why it keeps chats for 72 hours. It doesn’t explain why there’s no way to keep your history and skip training.
• What “anonymized” covers. The hub says your settings don’t control this processing, and doesn’t say more about it there.
• What Antigravity keeps with telemetry off. A user asked on October 11 how long data is retained with it disabled. That thread has no answer either.

Where Enjoy fits

Enjoy doesn’t run the Gemini app, so the Keep Activity choice is yours to make at Google. If you code with Antigravity in Enjoy, the Antigravity switch still applies. Enjoy has no AI model of its own and never sees your conversations. It hands them straight to Antigravity on your computer, and what Antigravity works on goes to Google under your account’s settings there (what AI providers see). Enjoy doesn’t watch or change that setting.

For work you’d rather keep on your computer, OpenCode with a local model runs in the same conversation, and nothing leaves your machine.

We’ll update this memo if Google responds or changes how the switch works.