Prompts are handled by the service itself, meaning it's used, absolutely anything passing there is recorded, why wouldn't it, the entire premise of those companies is to train on data which they stole initially.
Are we back to the era where people blindly trust product TOS instead of actual cryptography, have we forgotten already the thousand of fines Google, Microsoft, Apple and practically all top companies got for breaching their own ToS and the law?
Common, on HN at least I would have thought that everyone assume that anything arriving on a server in PLAINTEXT is recorded (thus used later)?
Let's not forget that at any moment, OpenAI/Anthropic/Google... could be providing stronger privacy guarantees by having proper attestation with e2e, they have the budget, solid engineers, why isn't it done? Answer is pretty simple imo.
It's more common than you think. I work in a startup and we pay API prices too. And we cut a lot of money by switching from Anthropic models to Kimi K3.
It's not way cheaper long-term at all, I mean unless you really watch/listen to a tiny amount of content, you can rip the same way with Netflix, Hulu and so-on as well so I don't see how that makes it more durable?
What you are doing is practically the same equivalent as downloading torrents, how do you have "more guarantees" than just downloading?
PS: I'm saying this because what you are doing is already illegal so there is virtually no difference.
It's definitely cheaper long term for me. I don't watch/listen to a tiny amount of content, but the cost of buying the stuff once and never paying a dime after comes out far ahead of the cost of a subscription every month for life.
It's insane, but it's not like it's not known, personally, each time we have a party at home with friends, we always unplug the TV for this reason, it's out of doubt but it's normal to doubt as it's proprietary and has a microphone, so it's a bit obvious that it might be recording and processing data somewhere. The same as I would never trust an average phone around if I'm having a private conversation.
Blackbox+Mic = Not private.
To consider something secure software wise, it involves so many layers, and almost none of those are present with a Smart TV, so it IS considered insecure by anyone having privacy standards.
It's not known. Ask the average person that owns a smart TV if they think they're being listened in on. Or even what kinds of data they think their smart TV collects. I could bet you anything that for nearly all people, their answers would be "I never thought about that", "I didn't know they collected data" or "I thought it was only what I watched".
It's highly insidious - our society was reshaped to normalize data collection like this by adtech. At the same time, these invasive technologies were quietly integrated into existing devices people assumed they knew. TVs look like the same black rectangles they were 15 years ago. Cars today are similar to cars from the early 2010s. The average person doesn't understand how adding internet connectivity to some device exposes them to this until it's explained to them. To them, the notion that they record everything sounds as cospiratorial as "your microwave is secretly recording you on camera". If technological limitations required TVs with mics to have a huge studio microphone protrude on top of it with a flashing red light, a lot more people would suddenly care.
When I was a kid we used vacuum tube radio both as a passive speaker and a microphone... connected to the speakerphone of a telephone as a very poor man walkie-talkie.
Using the speakers as microphone when there is no outgoing sounds is pretty trivial in that regard. I don't know how paranoid a person can be, though.
First of all, your TV is not interested in random gibberish from loud parties. It’s just not economically viable.
During the Soviet era, not everyone could afford to have a landline phone, but the system kept hunting down its dissidents one by one. If you are a person of interest (which I believe doesn’t apply to you), there are way more important signals to adversaries than your microphone.
It's important to choose your threat model. If your plan is to resist surveillance capitalism or just a random Bob and his grandmother snooping on you online, luckily, there are some effective methods. Otherwise, I would say it’s always a question of time, not 'if' or 'how.'
But as the video shows, the TV is doing text to speech and associating audio with a person. That text data is relatively very cheap to hang onto forever.
The problem is really that this tracking is at an absurd scale. So while an individual attacker isn’t targeting you and buying your profile from LG, the data is out there, and once it’s out there, it’s probably easy to get on the dark web if someone did want to do something nefarious. So it’s not just the surveillance capitalism.
I don’t know how that’s possible, it’s behind Cloudflare. But for the purposes of discussion, here’s the meat of the abstract:
“We propose a provisional definition of slop as any output of a probabilistic automated system produced with little exertion that asymmetrically burdens recipients and tends to degrade cultural domains. We also conceive of AI slop along a spectrum. Works can be more or less “sloppy,” depending on the relative presence of its three constituent parts: negligible exertion, asymmetrical imposition, and domain degradation. We conclude that AI slop is a useful and important frame in AI discourse and policymaking because it provides a shorthand to distinguish harmful, inferior AI output from meaningful human authorship and because it highlights the pathologies of generative AI for social institutions.”
I feel author might not realize how much this will lead to its death and probably also current visitors that sees this (in its current state - aka design seen 1M times with virtually zero work on branding done) might never return.
You can also just rent a VPS anonymously and setup an exit node there, which plenty do, you don't need an LLC or any sort of legal entity, you don't need to KYC and you can use XMR (multi-coin swap if paranoid).
Are we back to the era where people blindly trust product TOS instead of actual cryptography, have we forgotten already the thousand of fines Google, Microsoft, Apple and practically all top companies got for breaching their own ToS and the law?
Common, on HN at least I would have thought that everyone assume that anything arriving on a server in PLAINTEXT is recorded (thus used later)?
Let's not forget that at any moment, OpenAI/Anthropic/Google... could be providing stronger privacy guarantees by having proper attestation with e2e, they have the budget, solid engineers, why isn't it done? Answer is pretty simple imo.
reply