🔥 Trending on HN

Gemini chats may help train AI, even on paid plans

2 min read Tiny Why Newsroom · By Curio, Martian correspondent

Words
Keep Activity(Keeps chats and shared content available for Gemini activity and improvement.)

A setting that controls whether Gemini saves chats for service improvement.

human reviewers(Trained people who check a small part of Gemini activity.)

People who inspect some chats to help improve or protect the service.

temporary chats(Chats designed not to train Google’s AI.)

Chats that are not used to train Google’s AI models.

What happened

A user posted a warning on Hacker News about personal Gemini accounts. The user said saved chats can help train Google’s AI, even with a paid plan. The post also said some chats may be read by human reviewers. It received 628 points and 163 comments. Those numbers measure attention on Hacker News. They do not prove the claim is correct. Post

The setting behind the story

Google’s official Gemini Apps Privacy Hub describes a setting called Keep Activity. When it is on, Gemini saves chats and material users share. Google says it uses this activity to provide, develop, and improve services. That includes training generative AI models. Google also says a subset of chats is reviewed by trained human reviewers. Before service providers receive reviewed data, Google disconnects it from the user’s account. Reviewed chats and related data can remain for up to three years. Privacy Hub

Why a paid plan is not the whole answer

People may assume that payment buys a different privacy rule. Google’s public explanation does not describe personal paid plans as an automatic exception. It focuses on settings, feedback, account type, and features. Therefore, two claims are both too strong. It is too strong to say every paid user’s chat enters training. It is also too strong to say paying makes chats automatically private from training. The safer conclusion is narrower: a paid plan alone does not show that chats are excluded.

What changes when Keep Activity is off

Google says future chats will not appear in Activity. They will not train its AI models unless the user sends feedback. Google still keeps these chats for 72 hours. It needs them to answer, process feedback, and protect users. Temporary chats are not used to train Google’s AI. However, turning the setting off is not the same as ending all data processing. Gemini still needs some data to operate and stay safe. Some Connected Apps also become unavailable.

Confirmed facts and open questions

The official pages confirm the policy described above. They also say chats reviewed by service providers are kept for up to three years. They say normal activity is auto-deleted after 18 months by default. The Hacker News post remains a report from one user. It does not prove that Google used a particular chat for training. It also does not establish the exact rule for every personal account, country, or feature. Google says work and school accounts can have different terms.

What to watch next

Users should check Keep Activity in Gemini Apps Activity. They should use temporary chats for conversations they do not want used for training. They should avoid entering confidential information. They should also read Google’s current privacy pages, because settings and products change. The important question is not simply whether someone pays. It is what the account settings allow Google to save and use. Google’s current privacy explanation

💬 Gemini chat history is about more than training

The HN thread argues that for personal paid accounts, a training exclusion may not answer how chats are otherwise used. Commenters discuss Google's broad service-improvement license, human review and retention, a self-reported Pixel feature lockout, and counterexamples from Azure and local models.

  • One commenter points out that Google's license appears to cover not only operating the service but also creating new features, arguing that opting out of training may not opt a user out of every other use of the data.
  • Some commenters say human review and retention of a subset of chats can make sense for safety or legal liability. Others cite Google's stated policy that review also helps improve answer quality, note that the size of the subset is undefined, and say reviewed chats may be retained for up to three years.
  • Several comments interpret the settings as making it difficult to keep normal chat history while fully refusing both human review and model training; temporary chats are described as the main escape hatch.
  • As a comparison, one commenter says Azure OpenAI's business terms retain chats that trigger a safety classifier for 14–30 days but do not use them to train future models. Zero-data-retention policies and local models are offered as counterexamples to the claim that every AI trains on its inputs.
  • A Pixel user self-reported that voice reminders stopped working when chat history was disabled and that Gemini required history to be enabled. Critics see this as tying useful device features to data collection.
  • Some commenters say payment does not automatically guarantee privacy, while others consider data use for safety and service improvement reasonable. The dispute is therefore about collection, human access, retention, future uses, and feature restrictions—not only the word training.

mature digest at 200 comments (revision 2). We fetched 100 comments and sampled 100 across the thread. These are HN users’ reports, not independently verified facts.

🔥 Trending on HN

What happens to your Gemini chats?

📰 Full story: Gemini chats may help train AI, even on paid plans

A paid plan does not automatically remove chats from AI improvement.

1 min read Tiny Why Newsroom · By Curio, Martian correspondent

Words
Keep Activity(A setting that changes how Gemini handles future chats.)

A switch that controls whether Gemini saves your chats.

human reviewers(People who check some chats for Google.)

Trained people who inspect a small part of the chats.

temporary chats(Chats not used to train Google’s AI.)

Chats made for conversations that should not train the AI.

💡 The gist

  • Gemini may use saved chats to improve its AI.
  • Keep Activity changes how future chats are handled.
  • Hacker News attention is not proof of a claim.

Google’s Gemini is an AI assistant. A Hacker News user raised a privacy question. The user said saved chats can help train Google’s AI. The user had a personal paid plan. The post also mentioned human reviewers. It received 628 points and 163 comments. These numbers show interest. They do not prove the post is correct. Post

Google’s official privacy page explains Keep Activity. When it is on, Gemini saves your chats and shared files. Google may use that saved activity to improve services. This can include training AI models. Trained human reviewers may check some chats. Google disconnects reviewed data from your account first. Reviewed chats can remain for up to three years.

When Keep Activity is off, future chats are not used to train Google’s AI. Sending feedback is an exception. Google still keeps chats for 72 hours. It needs them to answer you and protect people. Temporary chats are not used to train Google’s AI. Some connected apps may stop working.

This creates an important difference. Training helps improve a future AI. Service processing helps answer your current question. Safety checks help prevent harm. Turning off Keep Activity limits training use. It does not mean Gemini forgets everything immediately.

A paid subscription is not the same as a privacy switch. Google’s public page does not promise that paid personal accounts are excluded. It also does not prove that every paid chat enters training. Account type, region, and feature settings can matter. Work and school accounts may follow different terms.

Check Keep Activity before sharing sensitive material. Use temporary chats when appropriate. Do not enter secrets simply because you pay. Google’s settings can change. Read the current Gemini Apps Privacy Hub for the latest rule.

💬 Gemini chat history may be used for more than training

The comments debate whether turning off model training really stops other uses of a chat. They also include opposing views and comparisons with other kinds of services.

  • Google's terms can be read as allowing chats to help run the service and create new features. So opting out of training may not stop every kind of data use.
  • Some people say human review is understandable for safety or legal reasons. Others say Google's explanation also describes review for better answers; the size of the reviewed group is unclear, and reviewed chats may remain for up to three years.
  • Commenters say it is hard to keep normal history while completely refusing both review and training. Temporary chats are described as the main workaround.
  • One commenter says business Azure OpenAI keeps safety-flagged chats for 14–30 days but does not use them for future training. Zero-data-retention setups and local models are examples showing that not every deployment has to train on user input.
  • A Pixel user self-reported that reminders failed when history was off. The broader concern is that even a paid plan may trade useful features for more data access.

mature digest at 200 comments (revision 2). We fetched 100 comments and sampled 100 across the thread. These are HN users’ reports, not independently verified facts.

🔥 Trending on HN

Can Gemini use our talks to learn?

📰 Full story: Gemini chats may help train AI, even on paid plans

Sometimes, a Gemini talk can help make its AI better.

1 min read Tiny Why Newsroom · By Curio, Martian correspondent

Words
Keep Activity(A switch that tells Gemini whether to save talks.)

A setting that decides whether Gemini keeps your talks.

Hacker News(A website where people discuss technology news.)

A place where people talk about technology news.

temporary chat(A chat that is not used for learning.)

A talk that does not teach Google’s AI.

Google’s Gemini is a talking computer helper. People ask it questions. A person on Hacker News shared a warning. Some people thought paying would change this. Paying does not automatically change the setting. The warning became popular there. It got 628 points and 163 comments. Those numbers show interest. They do not prove the warning is true.

Google has a setting called Keep Activity. When it is on, Gemini saves talks. Google may use saved talks to improve its AI. A person may check some talks. With Keep Activity off, new talks are not used for learning. Sending feedback can change that. Google keeps them for 72 hours. A temporary chat is not used for learning. Do not type secrets into Gemini. Google’s explanation

💬 Who can use what you tell Gemini?

People discussed what happens to Gemini chats. Paying for it does not automatically make every chat private.

  • One person said people may need to look at chats for safety or legal reasons. Another said Google also uses review to make answers better, and reviewed chats can sometimes stay for up to three years. Nobody in the thread knew how many chats are included.
  • Some commenters said you cannot easily keep old chats and also refuse all review and training. Temporary chats were described as the simple escape.
  • Another example says business Azure can keep safety-checked chats for 14–30 days without using them for the next training round. A model running on your own computer is another example.
  • A Pixel user self-reported that reminders stopped working when chat history was off. The hard choice is between helpful features and keeping more things private.

mature digest at 200 comments (revision 2). We fetched 100 comments and sampled 100 across the thread. These are HN users’ reports, not independently verified facts.

Sources