Tell HN: Gemini trains on chat history, even for subscribers (personal accounts)

3 days ago

If you have a personal Google AI subscription, all your saved chat history in Gemini can be used for training and there is no way to opt out unless you turn off chat history or use temporary chat sessions. These chat transcripts may even be read by human reviewers. Although Google claims to disconnect them from your account when human-reviewed, they explicitly warn not to enter confidential information.

https://support.google.com/gemini/answer/16836988

“[Gemini app activity] is used to improve Google services for everyone, including by training generative AI models, if your Keep Activity setting is on.“

https://support.google.com/gemini/answer/13594961?hl=en#human_review

“A subset of chats are reviewed by human reviewers”

https://support.google.com/gemini/answer/13594961?hl=en#pn_retain_data

“Chats reviewed by human reviewers [...] are not deleted when you delete your activity. Instead, they are retained for up to three years.”

https://support.google.com/gemini/answer/13594961?hl=en#pn_data_usage

“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.”

To disable their training on your chat sessions, you need to turn off “Keep Activity” in Gemini’s settings. Turning that off means you’ll lose any chat sessions after 72 hours.

I learned this too late, thinking that having a paid subscription = not being turned into the product. Apparently paying is not enough.

Thank you to this user on HN for mentioning in another thread: https://news.ycombinator.com/item?id=50355424

"These chat transcripts may even be read by human reviewers."

PSA: This is true of all the frontier model providers even if you disable activity tracking. They are all constantly monitoring your interactions for problematic behavior.

[0]: https://www.nytimes.com/wirecutter/advice/ai-chat-privacy/

[1]: https://www.404media.co/inside-project-lily-the-humans-readi...

[2]: https://www.winknews.com/news/woman-arrested-after-ai-threat...

  • The sole reason LLMs has gotten this good is because their aggressive expansion has been accompanied with literally this. The amount of uncredited information these models have been fed must be astonishing, also terrifying.

    • I think the heavily subsidized subscriptions are exactly for this purpose - the training data they generate is worth the high subsidy.

      7 replies →

    • Why do you think they’re getting into consulting now?

      Imagine all those tasty confidential IP and contract documents they have ingested. You couldn’t buy those.

  • That is, fundamentally the marketing ploy of “AI Safety”.

    For an AI system to be “completely safe”, like a self-driving car, it must map the full environment it operates in and be taught the “right way” to handle every conceivable scenario.

    So when you present it with novel data or a situation, that constitutes an “AI safety event”, which per their ToS is a carve out for training on your session.

    Further, go read the ToS carefully about what you opt out of. You’ll note there’s nary a mention of tool calls. You’re only opting out inputs (what you type) and outputs from the model from being trained on. All the content of the tool calls made while running on your system are fair game for training.

    • There is no way this is true, no? This'd mean literally all codebases people are working with have been trained on. There's no way anyone would use these guys if that was the case?

      5 replies →

    • Well for us Europeans, GDPR explicitly allows them to store pretty much anything if somebody can claim it’s for safety.

      It’s literally just complicating things and providing zero reassurance to us AI customers.

      4 replies →

    • Additionally, how stable are those ToS? Do they change every few days, like their models?

  • I can guarantee you that multiple people are reviewing some chats. This is very similar to Apen and whatever company prior to that for reviewing Google searches. They'd hire out contractors with NDA.

> Please don’t enter confidential information that you wouldn’t want a reviewer to see or Google

If Gemini has access to app data like gallery, mail etc that can also be subjected to human review? I imagine that will also be part of the context.

They are vague when they say 'don't enter'. Is it only prompts or does it also cover other data too.

  • Wasnt there an announcement some weeks ago here on HN that one of the BigTechs popular services are now including your personal pictures as well? Wasnt it Google?

  • Exactly, I have a Gemini Pro account, I can’t see how privacy can be maintained. If I ask Gemini to take the content of an email and added the dates to my calendar, then all of that private information might as well be public then!

  • Your email, Google docs, Google photos, Google home devices etc. are all subject to human review. The data is on their servers.

    And this goes for all cloud providers btw.

Is this really surprising to anyone?

I just sorta assumed that Google was hoovering up anything I type into its AI and don't tell it anything too personal or sensitive.

Is anyone under the illusion that our data is not being collected, whether it's Google or literally even the seemingly innocuous TVs in our home that have been found to be spying on us 24/7?

  • It can be difficult to connect the abstract knowledge that Google collects their data with the concrete reality of an AI training on that data, or a human manually reviewing it.

  • No, at least to me. However it seems there's no better option. The optimisc part is that those personal information won't be used by any others, though nobody guarantees.

  • People read privacy polices and believe them. It’s like EU cookies alerts that are automatically dismissed by 99.9% of people, accepting everything.

    • > People read privacy polices and believe them.

      Literally the opposite. People don't read privacy policies, even ones designed to be readable.

      The privacy policy says it grabs everything, and there's nothing that says they'll stop just because you pay them.

      2 replies →

  • Yeah, don't be under the illusion people, just take it and don't talk about it. It's not surprising, so just bend and shut up. It's not like you have courts and a Congress and this thing called democracy, you're helpless so why you keep being surprised at being cheated out and act like you can or want to do something about it, by starting with talking about it?

I don't trust any of the providers, but at the same time I think you're reading into the terms too much for wrong reasons. Yes, you have to opt out. But:

> A subset of chats are reviewed by human reviewers

This is expected these days for threads which may make the provider liable. Just look at the recent death case due to bad GPT medication recommendation.

> Chats reviewed by human reviewers [...] are not deleted when you delete your activity. Instead, they are retained for up to three years.

They're being retained for legal purposes.

> Please don’t enter confidential information...

This is so they don't accidentally preserve it for review, but also they don't want you to share private info that the agent will post somewhere without an explicit approval.

Those points make sense for them, regardless what other T&C rules apply.

  • What you describe is almost certainly what the legal team who drafted the terms and processes had in mind.

    But there's two reasons to bring cautionary attention to it anyway:

    1. As a user, I may not be okay with those terms as they stand and may not have the legal insight to have intuited them. In as much as I have a choice, because of those terms, I may want to use Gemini less/differently or look at other providers that have crafted less shrewd and defensive legalese and processes.

    2. As a executive or manager at Google, whether working directly in feature management, or advertising, or government compliance, these same terms define the framework for anything I want to develop. What "makes sense" from a defensive legal standpoint today can the open door for an exploitative or otherwise unexpected use later.

    I get what you're saying here in terms of plausible or justifiable "innocence", but that probably shouldn't get in the way of some critical personal reflection for users.

  • > This is expected these days for threads which may make the provider liable. [...]

    > They're being retained for legal purposes.

    It’s a significantly wider usage than that:

    A subset of chats are reviewed by human reviewers (including Google’s trained service providers) to help improve Google services. These include Gemini models, other generative AI models that power Gemini Apps, and technologies that help reduce unintended activations.

    Reviewers assess if Gemini Apps’ responses were low-quality, inaccurate, or harmful. They also suggest better responses. We use their suggestions to help our models respond better in the future.

    That’s in addition to the liability language:

    Chats are also reviewed to keep Gemini safe and secure for everyone and to address violations of the Google Terms of Service, including the Generative AI Prohibited Use Policy. Reviewed chats are retained for up to three years.

    https://support.google.com/gemini/answer/13594961?hl=en#huma...

    Some important things to take away from this:

    * They’re claiming to use chat transcripts for both training and for human review to improve training, not just for liability.

    * They don’t define “subset”. It could be most chats.

    * They make no promises of anonymizing content, just “disconnecting” it from the account.

    * This is happening for free and paid user subscriptions, by default, and the only way to stop it is to turn off chat history altogether.

  • You cannot opt out, except by using only temporary chats. You cannot keep your chat histories visible to yourself only. If you want to access your previous chats, you cannot opt out of human review and training models on your chats.

  • Azure's OpenAI terms are similar but more explicit with the safety and no-training terms.

    Some chats are saved for 14-30 days IF they trigger a safety classifier, but they are never used to train future models or retained after that period, the objective is to verify safety. Simple as that.

    If you are a business user, like in the case of Azure OpenAI, the analysis and priorities are quite different though, but the mechanism seems to be the same.

  • I disabled it because of the human in the loop.

    Don't care for what, store it away audit save if you have to, talk to laywers in case you need to read through them but don't let random people read my stuff.

    And in this case stuff = AI discussion which feels to me similiar to my google docs documents im hoping/expecting Google also to not read

  • > This is expected these days for threads which may make the provider liable.

    This doesn’t make sense. In such cases data could be reviewed by court order. That’s not something you’d have to agree to in tos.

  • I think such liability as "someone talks to the chatbot, gets wrong reply and poisons themselves" could be disclaimer by such conditions: "Our chat is unsuitable to discuss topics X, Y and Z. If you choose to discuss them anyway you're doing it at your own risk.

    But this would require an admission of limitations of the technology.

    Do people sue Google if search links to a page that has misleading info that results in someone taking a wrong drug? Nope.

    Do we sue the browser maker when the info on a given Web page is malicious? Also no.

    So, I don't understand why we treat chatbot providers any different. Everyone should be expected not to believe a word they say without corroboration.

    • That's like apples, oranges, and pears.

      You've narrowed the Google example down to them having a link in their index as a search result. People can already sue for keeping links to certain sites/data (pirate sites, personal data etc.) Somewhat closer for comparison, you could already sue Google if they misrepresented the links or content provided by the search, if it differed significantly from the site.

      You could already sue the browser for misrepresenting/modifying the info of a given web page, and/or injecting malware. They didn't create the website or content, we all know that.

      You can sue any company who is the creator/owner of information. Chatbots are technically creating new information, so whilst a grey area from purposeful human misinformation, the companies are far closer to owner/proprietor of information than the other examples.

imho you should act as if everything you type, not even send, is stored forever and might be used for training whenever they decide to retro actively change their TOS, no matter how much money you pay

I am not sure why anyone expected anything different from Google. This is a company that lives and thrives on user data and tracking users. Users are the material that is being run through this entire huge ad-selling machine.

  • Would you prefer to not have people discuss the topic, just because you’re already familiar with how evil google is? Lots of younger people didn’t have out experience of seeing companies evolve into the current evil they are, or do not follow the industry

  • I work with user data in Search and this is a common sentiment ... but I think it slightly misstates the purpose of using user data.

    The primary purpose has always been improving search overall -- traditionally this has meant updating query interpretation, ranking, presentation, etc for smaller and smaller query slices ("Ah, when we show the mayor and the population in response to [Barcelona] people just revise their query to [Barcelona FC], so we should make sure to include latest match result on the top. Not so for [New York]."), and running tests on every product change to see what people like. In this case nobody cares about any _individual_ user's data, only the collective. This is basically the original "collective intelligence" contract of PageRank, albeit hillclimbing on billions of searches and clicks instead of using link structure. (Old question: does original PageRank count as machine learning, or not?)

    Second is personalization -- things like "Oh, my flight or my delivery is expected on date X, handy that gmail popped a reminder on my calendar, good work Google." Or searching for a restaurant in Maps and seeing results that are biased towards the kinds of restaurants you already like, that kind of thing.

    When it comes to revenue the purpose of these isn't to target ads, it's just to have you get value from and use the product, and when you are ready to spend money this is where you come to search for the thing, whatever it is.

    Finally there is limited use of a limited set of data for ads relevance. There's higher CPC for ads you might actually be interested in (and if you are looking to spend money, is finding a willing seller who has compelling terms via an ad such a bad thing?) But ad revenue is mostly downstream of users being satisfied with the core product and using it more, whatever that product is. (Search, Maps, AI chat, etc)

    (Bonus. My Search pro tip re: ads is leverage the modern stack and spell out your intent with a few more words when you query. We're pretty good at not serving ads for non-commercial queries. But old-school super short "searchese" queries make it hard to guess whether your intent is commercial or not.)

    • > The primary purpose has always been improving search overall

      That is what I believed, a long time ago, too. At this point I'd say this kind of thinking falls somewhere between idealistic and naive on a spectrum.

      If better search were indeed the goal, we would have had better search a long time ago. But we don't. In fact, search gets worse over time, devolving into an ad-ridden mess freely gamed by SEO content farms, and now with a dumb-but-cheap-AI cherry on top.

    • Oh, I get it now. The anti-user behaviour is actually there to help me as a user. The main goal isn’t to Hoover up all my data to profile me and serve more targeted ads in more places. The main goal is to make search better. Now I get it.

      1 reply →

The implicit contract used to be simple: if you pay for a subscription, you aren't the product anymore. its broke!

This is true of all paid “Pro” type plans from all the AI companies. These plans fall under “Consumer” style plans that do not explicitly forbid training. This is inn contrast to actual “Commercial” plans that do have formal Data Processing Agreements in place.

I respect Google’s position on the language in the Consumer ToS. It’s plain as day that they train on the consumer plans.

https://antigravity.google/terms/

Section 5, in Bold. “We use Interactions to evaluate, develop, and improve Google and Alphabet research, products, services and machine learning technologies. Interactions are collected and used in accordance with this Agreement and the Privacy Policy. Google employees and contractors may access, view, review and use Interactions. Please don’t submit confidential information or any data you wouldn’t want a reviewer to see or Google to use to improve our products, services and machine-learning technologies. If you don’t want your Interactions used in this way, navigate to settings to change your preference on how such data is used. Even when your Telemetry is toggled off, Google still uses your data to provide the service, process any feedback and help keep Antigravity safe.”

That last line is clear that they “use your data to provide the service”.

This is in contrast to the Google Workspace ToS, where it clearly says they don’t train (but it’s really hard ti find:

https://workspace.google.com/terms/service-terms/

“12.11 Training Restriction. Google will not use Customer Data to train or fine-tune any of its generative artificial intelligence models supporting the Google Workspace Generative AI Services without Customer's prior permission or instruction.”

Of course, Workspace doesn’t give you a good coding model, for that you have to go to API billing, which also has the DPA attached.

Anthropic and OAI have similar Team/Enterprise plans that prohibit training also, but they cost more than the basic Pro plans.

While we're touching on de-googling.

Can anybody recommend a good Swype¹ keyboard that is not G-board?

Old Swype+Dragon is dead.

Google regularly removes words they don't like from my dictionary. For example swear words, or today, I had to re-type GrapheneOS.

AD 1: By Swype I mean writing on Android by dragging my finger through the letters making up the words I want to type.

I also miss being able to easily use ctrl+c,x,v or easily entering space or other special characters, using special gestures.

There is just so many FOSS keyboards, that it's hard to choose from them. And I hope somebody with similar usage patterns could share what they are using.

Even if they don't train on your content, Google reserves the right to use it for improving their services in other ways:

“ This license is for the limited purpose of operating the services, which means allowing the services to work as designed and creating new features and functionalities. ”

https://policies.google.com/terms?hl=en-US#toc-permission

The training exclusion has always been a distraction. These service providers have plenty of other uses for your data.

Is anyone actually surprised by this? They've devoured the whole internet for training already, the stream of shit coming in from users is the only way to hoover up more training data.

  • Yes IF the Terms of Services and UI communicate otherwise. It's not normal for a company like Google to break ToS or Privacy Policy, would be quite novel.

It's good to record this sort of information, but I hope no one expects any of the big LLM tools to be even remotely private. This is big tech in 2026. They don't even need to prove they can make money on your information. They're going to take every little last bit of information they can conceive of. 100% of the time.

Yes, I wpuld be surprised if it was not. The type of information people are putting into these models, with blind faith people aren't reading them, is mind boggling really. One reason locally hosted models "win" is things like "sort out my monthly finances " and stuff you do not want to be saying to Big Tech.

This has an easy solution. Use Google Workspace, their private policy there promises to not using your chats for training. They don't want to alienate companies, so their privacy policy is much stronger there.

It's stuff like this that makes me happy that at least one big tech company (Apple) cares about privacy. Hope they don't lose that. Although I wonder how much of that is exactly why they're behind in AI...

They should offer history + not training (every other big AI company is offering that). The only way to get that currently is to use Gemini in Google Workspace.

There is also no opt out of training on their personal AI coding side with Antigravity.

And while there at it they should also allow to buy Youtube Premium Full normally when you've got Google AI Pro yearly which includes Youtube Premium Lite. The add+on price is like over 100% the price than buying Youtube Premium Full in my country and no way to buy normally :(

I know it and use Gemini based on this background, with another background that most of my personal information is not interesting for potiential human reviewers. Just take it easy ~ :-)

For subscribing users: Disable “Keep Activity”, don’t use Gemini Web/Mobile apps and do all your chatting needs inside a Gemini Notebook. I understood that Notebooks are walled off. To the extent that that is true I cannot be sure. YMMV. Alternatively, maybe run models locally.

Does this mean that if I take the time to correct an AI and show it how its wrong, later training data will include the correction?

That would be quite good actually.

Right now when the AI wrong I don't bother to correct it, I just open an new chat and change topics.

  • I guess that leads to the next question, how does Google filter for quality when including other user chat? They wont want the training poisoned by all the dumb shit people argue with AI about.

This shouldn't be surprising at all.

Anytime you aren't paying full price for something, you are paying the rest with something else. We know the real cost because there is an API that comes with all the confidentiality, and that is much more expensive.

FYI, this is true for the other labs too.

It might matter if Gemini was good enough to use but its not - I only use ChatGPT and Claude.

  • For daily chat gemini is great.

    actually even gemma locally is great for a lot of chat purposes.

  • Their recent Argon is weak as well. But then they promise Carbon going to be world changing. Yeah sure.

Is there a way to use an API to export locally all the temporary chats and restore them as temporary chats for the next session? I am sure they’ll disable those eventually too, or even worse block your whole account for misuse.

What can one currently use instead to have the same experience but preserving privacy?

I assume that everyone does this no matter what they say. Seems like too juicy a target

  • I wouldn't go that far. These companies are leaky and there can be a backlash that causes people to lose trust and move to competitors. For instance, when people discovered the Grok CLI was uploading repositories to the cloud despite the training toggle being disabled.

    • I knew grok was garbage from the start and that I couldn’t trust google. Anthropic I’m still not sure but we’ll hear about the dirt when they ipo

  • Obviously anthropic are doing this. Hence their "don't insult Claude too much" policy.

I know they say not to enter confidential information, but they are basically using the LLM to spy on users it's unfortunate. Even deleting chats doesn't help because they keep that information for another year or 2.

In this true age of AI, privacy is virtually non-existent. I now make it a point not to share my secret keys or account information when chatting with Gemini, for fear that such data might be leaked.

Why do you all think they got so good at coding and why they give us cheap personal cheap subscriptions for $200/month?

Of course they have been using our data all this time.

The fact that HN has to be told this is quite surprising. Google et al have been doing this underhanded, sneaky behavior well before LLM hype.

Privacy at big tech is a facade.

This probably explains why I see random google employees I've never met in my facebook "people you may know" feed.

I use File -> Save Page As ... does a good enough job, I just have to keep a folder on my computer.

that's why i don't use gemini for anything serious or personal. if you want persistant memory you should enable keep activity, but keeping activity allows google to use your chats.

  • I don’t use Gemini at all. It also gives me sub optimal experience compared to other personal assistants

I’m disappointed it is not common knowledge that:

1- In Google/Gemini/Antigravity there is no “opt-out” (without making the service completely unusable). Your code/data is directly in their training set.

2- Others have partial “opt-out”. They can still read your chat history, run models on them, probably indirectly train some internal models.

It's just keeping track of all the times you were mean

  • I sure hope so. Gemini is useless garbage, and every time I'm mean to it is a case in point. It's by far the most frustrating of any LLM I've ever used.

What's worse is if you have a Pixel phone then Gemini is completely useless without enabling this setting.

I had it switched off:

"Hey Google, remind me at 7pm <something>"

"OK, your reminder is set for 7pm"

Next day:

"Hey Google, why didn't you remind me at 7pm"

"I can't do anything with your phone unless you enable chat history."

WTAF!

  • This was the case with Google Assistant too.

    Google has this pattern of all or nothing approach.

    Maps refuses to show your *current* location if you don't enable location accuracy i.e. if you won't allow it to spy on your WiFi or Bluetooth it refuses to show your current location at all. GPS, what's that?

    We can disable notifications and even phone access but physical activity is the one permission that cannot be disabled for play services !?

    Google apps have a lot of WTFs!

    • And the YouTube home page is blank if you have Watch History off. The recommendation algorithm still runs on the side bar of any video you watch, but the home page refuses to even show your subscriptions.

      16 replies →

    • > We can disable notifications and even phone access but physical activity is the one permission that cannot be disabled for play services !?

      Sandboxed Play Services works on Grapheneos without physical activity so it's definitely not actually required by their services.

      But I think they showed the best example of that behavior just recently: no AI, no Google search for students because search is "too integrated" with AI (https://www.forbes.com/sites/maryroeloffs/2026/11/03/google-...)

    • I've just turned off location accuracy, google maps displayed a popup said results will be better if I turn it on, and after dismissing the popup it did use GPS to localize me with no issues.

      6 replies →

    • Surely that approach is implemented to force people in to the all group.

      Then again, anon YouTube no longer showing recommendations on the start page is what waned me off the website.

    • Last time I tried, Maps also refused to show a map at all, from anywhere if you didn't enable your location.

    • > Google has this pattern of all or nothing approach.

      Call it what it is, please. User-hostile behaviour that compels you to turn their tracking on if you want to use any of the practically helpful services.

      I'm a super privacy conscious person, and even I end up keeping many of these things on simply because I want to use for eg google assistant.

  • Worst part is that the history lives with your other gemini chats. So, "set timer for x" 85 times gets muddled into chats where I'm doing deep research.

  • Yeah, I am hugely annoyed by this. I kept paying for the subscription for a year and a half, but I think I will stop now, as I feel that the added value is very small in comparison to Claude Code or just using openrouter open models.

  • This is the most infuriating part. Complete bait and switch by bricking features of the phone I bought before Gemini for practically anything that used assistant.

    They are not transparent about it either. Anything you link to it to enable basic functionality you had on device before gets them the ability to use it for not just training but ads and measurement.

  • This has been annoying me too, and I was getting pretty close to turning on chat history having forgotten why I disabled it.

    Thanks for the reminder OP!

  • But don't worry, Gemini is nearly completely useless even if you do enable the setting.

  • Yeah its the worst. I reported it as a bug to them. Obviously ill never get a response

  • This is a risk issue. They need users to explicitly opt in or they risk user-data-related compliance suits. This has always been Google's problem, as an advertising company, and why they've taken so many seemingly awkward paths over the years with both Android & Cloud.

    • Or they could just not collect the data? Apple's assistant may suck, but it degrades gracefully if you don't grant it certain permissions.

And if you do want privacy, and you purchase a workspace business plan instead… well then you have the pleasure of paying more and not having access to the latest models (and having no idea when you might get access to them eventually months later).

This why I advocate Deepseek, GLM, Kimi, Qwen as in hosted directly (send data) to China. Think about it, what can they do worse off than what our own backyard AI can do? There is no privacy advantage in Western AI. It is way more likely ours leak your data to dark web (solid track records for decades) or get law enforcers on you. Meanwhile, am I suppose to be worried, China sending their Ming-like Eastern Bureau to my home in the west? On dark web, aplenty western companies leaks. How frequent the eastern one?

  • That’s a take I see often, and it always overblows the direct personal threat while understating the more abstract societal one. A Western AI company might report you for plotting a murder or a mass attack, but that’s not the issue for most people. Mass observation of what society thinks deep down and manipulating it with that data is a problem for both, and maybe more so for external actors, since they don’t have even a remote reason to care about your well-being.

    You’d maybe want to counter with something like 'But Google/Amazon/whoever just wants to sell me stuff!'. What makes you think that Chinese B2C would just pass on that opportunity?

    • I feel that saying "they just want to sell me stuff" undersells what Western data markets do with our data. It can and is used in ways beyond just advertising.

      I also don’t know that I agree with your idea that manipulation is a greater risk from foreign actors. You're right to point out that it can be a problem from either, but local actors have a particularly direct incentive to shape the opinions and behavior of the population whose views affect their own power and interests. I would not argue that means foreign actors are harmless, I just don't agree they are necessarily worse either.

  • The soviet union collected data and profiled American citizens so that they could (potentially) be used as dumb pawns if needed. Dumb pawns that didn't even know that they were doing the bidding of the soviet union.

    Just like how Kim Jong Un killed his brother using dumb pawns to actually carry out the murder.

    So if your head has not grasped this reality, then they are already two steps ahead of your "What could they possibly do".

    This is also the reason why congress keeps banning Chinese enterprise. It's not because they care about privacy, it's because they care about people being turned into pawns.

They’ve been doing that since Gemini was released!

I’m surprised it’s only trending now, I guess nobody reads the ToS and privacy policy :p

I always keep this in mind when interacting with proprietary LLMs. As someone who heavily relies on residential proxies over Tor and constantly rotates accounts (or avoids creating them entirely, as with Gemini Flash), I sometimes exploit this to subtly inject or promote public-interest ideas (mostly related to privacy) into the training data.

That's AI. Who thinks their data is safe with AI companies? That's literally their business.

training on "no i don't need google search to nurse me after a breakup, i was just asking what movie has the line she just dumped me"

You guys must be joking if you don’t want the AI to get better stop using it. NIMBY for your precious chat logs.

This is one of the reasons why I unsubscribed from Gemini Pro, apart from the fact that the experience with Gemini has been frankly underwhelming.

  • May I ask where you subscribed instead and why it is better and more secure there?

    • Well, I know it's likely self-delusion, but ChatGPT has an "improve models for everyone" switch event if the history is not switched off. Plus, at least it doesn't forget the context of the chat every 2 messages. For now.

Friendly reminder that you should take measures to ensure your agentic workflows are secret-blind. E.g. place secrets by reference, check by reference, do NOT have agents read secret sensitive files without redaction measures in place, and be prepared to rotate on contact

it’s all trained. that’s why i made trustedrouter so that it can be end to end encrypted ai chats

If a company can do something that benefits them, and there is no way the customer can find out or prove that, the company will do it.

Every AI trains on the input given to it

  • This is false.

    eg Gemini enterprise has a note under each chat session explicitly saying it’s not used to improve the models

  • ZDR policies are a thing. Maybe you can't trust all of them, but when a provider like AWS says it, you can probably believe them. They'd be setting themselves up for an endless stream of massive lawsuits from other large companies if they were lying about it.

    Local models are also a thing. The Qwen 3.8 I'm running on a spare computer is definitely not training on my inputs.

we already knew this - google is gonna google - same way they train on your google photos and emails.

I mean, this is google we talking about

also the same cases for antrophic and open ai

why would anyone trust company that literally need data to improve their product ????

>thinking that having a paid subscription = not being turned into the product

This has never been true in techbro capitalism land. The paying members of a project are even more valuable in your datasets because you want to understand their behavior so you can pull more whales, I mean convert some customers :)