LLMs killed the privacy star, we can't rewind, we've gone too far

[Rusty Snippets]

You'll find these days that there's no hiding place​

Add privacy to the list of potential casualties caused by the proliferation of AI, because researchers have found that large language models (LLMs) can be used to deanonymize internet users – even those who use pseudonyms – more efficiently than human sleuths.

Much of the academic work on online privacy over the past 25 years builds upon Latanya Sweeney's 2002 research on k-Anonymity [PDF], and prior research in which she demonstrated it is possible to identify 87 percent of the US population using three anonymous data points – a five-digit ZIP code, gender, and date of birth.

The possibility of identifying people from anonymous data became one of the central concerns about online advertising and the usage of cookies in web browsers.............................

AI takes a swing at online anonymity

: You'll find these days that there's no hiding place

www.theregister.com

 

[Curious Wonder]

Do you think this will end badly?

We've had cookies for years and nothing seemed to have happened.

 

[Anthony Parsons]

There has never been a hiding place, that is merely an illusion. When I was in the Army decades ago, the stuff they were doing opened my eyes. Nothing we did was private, phone, email, text, nothing... all private devices, if on operations, exercise or around a military base, they accessed everything. Now... I couldn't even fathom how much Government see. VPN's are an illusion for privacy... they only help hide you from corporations advertising cookies and such nonsense, nothing substantial about privacy. Read what Facebook and Google TOS, what you give them access to using their services. It's mind blowing what people are so easily providing about themselves.

 

[qubn]

Privacy was never real. Anyone who's run a forum for more than a year knows this, you can see IPs, emails, posting patterns, device fingerprints. Your users were never anonymous to you, and they were never anonymous to their ISP, their government, or any ad network with a pixel on your page. Pseudonyms don't help when your writing style, posting times, and topic interests are a fingerprint on their own.

LLMs didn't kill anything. They just made what was already happening faster and cheaper. The Sweeney study is from 2002, that's 24 years ago. If three data points could deanonymize Americans back then, what do you think a decade of social media oversharing has done to that number?

Saying "privacy never existed" from military experience is right, but for the wrong reasons. It's not about government surveillance, it's that people voluntarily hand over everything. They post their face, their location, their workplace, their kids' names, their daily routine, and then act shocked when an AI can connect the dots. And to the cookies comparison, cookies tracked you across websites. LLMs can infer who you are from the content itself. It's not the same league.

If you actually care about your users' privacy as a forum admin, stop philosophizing and do something practical: block AI scrapers aggressively, put content behind registration walls, strip metadata from uploaded images, rate limit crawlers, and use Cloudflare's bot management.

 

[smallwheels]

Privacy was never real. Anyone who's run a forum for more than a year knows this, you can see IPs, emails, posting patterns, device fingerprints. Your users were never anonymous to you, and they were never anonymous to their ISP, their government, or any ad network with a pixel on your page. Pseudonyms don't help when your writing style, posting times, and topic interests are a fingerprint on their own.

LLMs didn't kill anything.

The title of the thread is the title of the article on threregister and refers to the song "video killed the radio star" which was the first video ever played on MTV.

Saying "privacy never existed" from military experience is right, but for the wrong reasons. It's not about government surveillance, it's that people voluntarily hand over everything. They post their face, their location, their workplace, their kids' names, their daily routine, and then act shocked when an AI can connect the dots. And to the cookies comparison, cookies tracked you across websites. LLMs can infer who you are from the content itself. It's not the same league.

True what you say about the users providing their data, but most of them are not at all aware of the consequences. In fact the advertising industry has created kind of a mass surveillance system and is trading the data. So wouldn't it be more adequate to point at them and asking, if this is a business model that a society is really willing to accept instead of victim blaming?
Art from that it is really not a question that with AI there it is now faster, cheaper and easier to connect the existing dots.

If you actually care about your users' privacy as a forum admin, stop philosophizing and do something practical: block AI scrapers aggressively, put content behind registration walls, strip metadata from uploaded images, rate limit crawlers, and use Cloudflare's bot management.

Of your five points I lack Cloudflare (b/c I don't want it - ironically for privacy reasons) and partly the image metadata stripping. The rest is set. Oh, and I don't run ads on my forum.

Edit: Personally I'd consider it a bad sign for a society if you, if you don't want to be completely commercially exploited and tracked, have to care what you say, what you write and basically have to stay off the internet altogether. This would be a very harsh limitation. Personal freedom for me includes the right for privacy and if a highly commercialized society endangers that through some business models we should talk about whether these business models and practices should really be allowed.

 

[Alpha1]

The implications for forum users and forum admins are far greater than anyone in this thread seems to realize. I'm not going to spell it out, but once privacy watchdogs or forum users catch on, it can be quite devastating to forum admins.

 

[smallwheels]

The implications for forum users and forum admins are far greater than anyone in this thread seems to realize. I'm not going to spell it out, but once privacy watchdogs or forum users catch on, it can be quite devastating to forum admins.

Absolutely true.

 

[Alpha1]

In time, we might need some masking functionality so that member posts can be reworded and can no longer be used to identify forum members.