Yes, yes selfhosting is the most privacyrespecting way, I know, but I'm not a richfag. So here is my idea:
I use something like lambda.chat with a VPN. Is this enough, to prevent them from creating a shadow profile of me? Or should I stick with duck.ai ? I like lambda.chat more, because it has Dipsy. What do you think?
how about just don't
You should also copypaste your complete inputs from a notepad and change your writing style each session so they can't fingerprint your writing and input styles. Also use different browser and devices/VMs if possible. Actually, they can just fingerprint you based on what you tell the LLM. Basically
>>29195>>29195>>29196By that logic I shouldn't use the internet at all and it makes no difference, if I use online AI with some privacy enhancing methods or I just use ChatGPT with a regular account.
Nope. I know privacy is a difficult game, but I also know, it isn't all or nothing.
>>29197Even using a VPN like you said will narrow you down to the "VPN users" group of people. By trying to achieve privacy (presumably, you have researched VPNs, Qubes OS, etc. online before) and posting here no less you have essentially signed a warrant to get all your traffic intercepted, checked and even filed if deemed worthwhile. And if a government hasn't got you on their radar then you have certainly been tagged by corporate advertising algorithms. Or maybe not, who knows. But you will get profiled regardless, posting "Nope" won't help it, especially in the case of a chatbot that scrapes all available data to train the model.
>>29198But, you should do it anyways because it is funny to hinder anyone trying to do profiling
You can at least use ChatGPT on Tor with a disposable e-mail.
>>29198So you are telling me, that I should just do nothing and privacy is a meme. This is exactly what a corpoglowie would say.
>>29211Feel like privacy is the idealist counterpart to materialist operational security.
Privacy research can also be good fun and useful as training for actual operations.
The question in this frame is what material consequences are you attempting to avoid by implementing certain security practices.
Say this as a bit of an idealist, but not a privacy advocate. >>29212Basically, threat models.
>>29212I'm not Edward Snowden and my goal is not ultimate privacy. I'm not even willing to pay for privacy, I am not buying a Pixel phone just to install GrapheneOS. I am using a generic android device and I know, in the view of these privacy dogmatists this is a cardinal sin. So what is my goal? I simply don't want to give these corpos EVERYTHING. When I have the opportunity to reduce my digital footprint with no extra cost, I will do it. Total privacy is not my goal. I believe, this "all or nothing mentality" in these privacy communuties is dumb and counterproductive. And I certainly don't see, what this has to do with "idealism".
>>29214>I simply don't want to give these corpos EVERYTHING.You don't want to be manipulated, or you're looking at ethical-consumerism-tier praxis?
If you don't want to be manipulated then it's easier to install an ad-blocker and use privacy respecting services.
Or even rather weak things like disable search history in youtube etc.
>And I certainly don't see, what this has to do with "idealism".Well the idea was if you were pursuing privacy for its own sake then it'd be just a psychological or ideal notion without a material basis, e.g. don't want my landlord or boss or the FBI to throw me out because of posts on social media.
>>29215It can also just be good fun, but even the illusions might be part of the enjoyment.
>>29211No and I even posted this afterwards so you would not think so
>>29204But it is funny regardless that you think you can use LLMs privately those are pretty high on the "I will steal all your data" list even if the VPN itself doesn't leak anything, or nobody has tapped the wire on either side of the VPN and can decrypt it's traffic. It will propably keep out generic advertisers in any case.
>>29204If you don't know what you're doing you might make yourself stand out more than the average person.
>>29220Likely already happened as there are only so many people out of many billion humans that have been so concerned for their privacy as OP
https://www.ndr.de/fernsehen/sendungen/panorama/aktuell/nsa230_page-1.html Ok, how about this: How is using a online AI via lambda.chat (no registration needed) in any way worse from a privacy perspective than posting on this site here? We don't know who is actually behind leftypol. How can we be sure, they don't have a secrete deal with databrokers or a LLM company as well? People here who tell me I shouldn't use online AI, don't have the high ground here.
>>29215Privacy for me has a material basis: I don't want to give my data the corpos/I don' t want to give them even more power.
1000% materialist reasons here.
>>29236It is a valid observation that both could potentially be equally insecure in which case you would be leaking data twice as hard by using lambda.ai and leftypol. I don't know how that would change the value of someone's opinion. But leftypol is atleast marginally more secure.
You can use .onion and there's some basis to believe this site's not completely sellyourdata-based or a honeypot, your info is not being outright stated to be sold and used unlike on chatbot sites. And you hopefully don't leak info willingly about yourself here expect that you are a leftist extremist, but you might use a chatbot for porn which is very incriminating. You could even argue that when using a AI sites, you will be connected from the same address and setup for a longer period of time than when using an imageboard (and prompts/models/settings you use are easier to fingerprint than the couple short replies one might make here) which makes them less private in comparison.
>>29237If you agree with ethical-consumerism, or similar initiatives ("be the change you want to see in the world") then see no reason why you might not also take action to oppose the creation of these (privatized) models or statistics.
It's plausible this isn't exactly the "change you want to see in the world", (e.g. you want to see a continued advancement of the means of production etc.) but this is just a shifting of the goal post, and you're free to do as you please.
For the record my previous suggestion
>>29215 would just imply to use
https://duck.ai maybe with the llama model. And will go ahead and switch over to this myself now.
I use Copilot through Tor and without logging in
GUYS CHECK THIS OUT!!!!
https://app.unlimitedai.chat/COMPLETELY UNCENSORED!!!
>>29286Alright bois and girls, can somebody explain zo me, how such a service can actually exist? How do they make their money, since there are no ads? I don't think they use the data for training, but it might be possible they are selling the data in some way. Give mr your theories! It is certainly an interesting business model.
>>29316They just lose money, and get continually pumped up by investors.
There are paid for plans, and for API access but apparently even the plans lose money.
They keep getting insane evaluations that allow them to do this:
https://www.cnbc.com/2025/03/31/openai-closes-40-billion-in-funding-the-largest-private-fundraise-in-history-softbank-chatgpt.html >>29316https://www.whois.com/whois/unlimitedai.chatSend them email if you want to ask. I think it's a company trying to get users fast, explains the many hits of chinese websites that essentially mirror their FAQ as well.
>>29317what a stupid and irrelevant post. the whole thread you behave like a bot or a glowie. Man, this place is so rotten and dead…
>>29319Didn't realize what thread was replying in, sorry.
It's somewhat related to the comment from
>>29239>It's plausible this isn't exactly the "change you want to see in the world", (e.g. you want to see a continued advancement of the means of production etc.) but this is just a shifting of the goal post, and you're free to do as you please.But it wasn't quoted.
>>29286Found some more info on this project. Here are the system prompts:
https://linux.do/t/topic/576697 Unique IPs: 12