That remains to be seen.
Nevermind. I tried it without the WebSearch feature, and it said it had hard limits on such speculation about a real person. It seems to be unaware that I could just click a button to give it the ability to also search the web.
All of its stuff about me is rather outdated though.
Edit: I asked it to look at my more recent stuff. It hallucinated that I had changed my name to YWTF, of all people.
I’ve had a bad day, and that made me laugh.
To be clear: The original post with the (apparently vague) information was publicly-visible (I edited that to remove the identifiable information). Another poster saw the information, and sent a PM to the first poster. That poster then wondered if they were out of line, and so also sent a PM to me to ask.
At that point, I still couldn’t see the PM in question. Theoretically, there must be some level of administrator who can read PMs on a whim, but it’s not us ordinary moderators. The only way we can read PMs is if they’re flagged (which, yes, is something that you can do), or if one of the existing members of the PM invites us to it (yes, multi-person PMs are also a thing, and others can be invited to or removed from them). In this case, the second member invited me, so I could read the PM thread.
From what playing around I’ve done with it, it has only ever skimmed little bits of posting history. It can’t even skim a whole thread well for me. The way the page is structured or something, it has difficulty “scrolling” through the thread to read more than a handful of posts. I’d be highly skeptical that it can read through all of your posting history without a lot of handholding at this point. The best I’ve been able to do is upload my own posting history (we have the option here to download all our posts as a spreadsheet file), and even with that, it had some difficulty keeping track of who was talking when in threads (it mistook a lot of quoted replies as my posts, but with additional prompting was able to sort most of it out), but it obviously had a much cleaner access to my history and was more thorough and detailed in its responses to queries about it. It even helped me dig up some posts I had forgotten about or had difficulty find. (One such query was for it to find the most emotional I had been on this board, and it found exactly the thread I was thinking of right away – one about a girlfriend cheating on me. Not particularly over-the-top emotionally, but out-of-character for me on a message board.)
But with a bare query without feeding it the actual archive file, it couldn’t find that, because it just doesn’t work that way.
While I fully agree w the thrust of the OP, here’s a point we seem to be overlooking.
No matter how carefully I never post anything going forward that is related in any way to me IRL, 100% of what I have posted to date is still there for the picking. And always (?) will be.
Further, even if TPTB shut down and deleted the entire SDMB 30 minutes from now, every entity that really wants to identify us already has the entire 26 year history of SDMB in their can.
The already-cited thread about what any AI knows about doper-you is sobering. But I don’t know how much of that is driven by current at-the-moment searching and how much is general background knowledge already baked into the AIs. Given the AI’s response time to queries, I suspect a lot of it is already “known” to the AIs’ own ache.
Next point:
Substantially everything valuable about SDMB comes from us sharing things about ourselves. FQ lives on the job- or hobby-related practical knowledge each of us brings. Anyone who discusses a topic knowledgeably is giving away clues about who they are. Ditto IMHO for advice on mundane household tasks, things to see when traveling to X, fashions, etc.
In the world of military security you’re trained to never try to “talk around” a secret topic hoping to convey real info to somebody else on the inside over a non-secure channel. The mistaken hope is that only the person you’re talking to has the context to make sense of your sneaky oblique talking-around. Any eavesdropper will just be bamboozled.
The reality is the enemy gains from hearing your context and with only two or three people talking around, the overlaps and collective omissions paint a pretty clear picture of what wasn’t said and why.
As applied to the Dope, that says anyone who says anything about anything is slowly and faintly inking in a target on their forehead, stroke by stroke. Each post is a form of talking around. But with enough posts, the resulting target is big, bold, and starkly black and white. And that requisite post number is very small. Far smaller than the posting history of most of us.
I’m not here to say anyone, much less everyone, should immediately withdraw from the Dope. Nor am I petitioning Ed to shut it down and delete it ASAP.
But we’re each in a hole. Some (much) bigger than others. And any material effort by any of us to stop digging could have a very chilling effect on the viability of the Dope. IMO we’ve been flirting in the neighborhood of minimum critical mass for awhile. I’d hate to see us crash over this.
I asked chatGPT about solost, and at first it said it couldn’t find anything about me. It said it’s most successful at finding info on high-profile, recent posters. Was I sure solost was spelled correctly? Kind of hurtful, actually 
When I pressed it to dig deeper, it did supply a profile of me and my posting style, with guesses as to my general age group, interests and location, which were accurate enough. It did not make a guess as to my actual identity, saying even if it could tell from my posts, it would be doxxing.
So liars and trolls are safer from doxxing than honest, upright folks. Good to know.
I say this tongue in cheek, more or less. At this point, at least, my point is that AI doesn’t seem to be able to tell the difference, either in what it reads or what it regurgitates, between truth and fantasy or falsehood.
“‘Truth’? What is that? This does not compute”
I haven’t tried my username x SDMB, but last night I was looking something up and the top o’ the search AI said I’d written a book I’ve never heard of. Score?
I was the poster who got identified by ChatGPT.
I was surprised AI could do that - but now I know it can. 
Both the poster who identified me and Chronos behaved impeccably. 
(I was initially upset until I realised the poster had sent me a Private Message (i.e. didn’t reveal me to the board) and that my original post has been smoothly edited.)
It’s been a useful learning experience. 
I’d prefer not to have my pseudo-anonymity pierced, but honestly it becomes less costly as the more intemperate remarks are put on record by others on Facebook et al. To some extent I don’t have to outrun the bear, I only have to outrun my fellow netizens. It’s better to lose your privacy after almost all others have lost theirs. Obviously you should keep exploitable specifics such as account numbers to yourself - that’s another level of bad.
I don’t think there’s any solution, period, easy or hard. For quite a while now, I’ve felt that “privacy” is like “carbon paper”, I know what it is and what it means, but it has no relevant meaning in 2026.
Yeah, I’ve had this approach since around 1999. Hell, I would send potential dates links to the SDMB, my Twitter account, etc., and I once uploaded a graphic of my college info which had my full name, etc., to this very site. Hell, here it is again:
I use my real name in here, so I’ve got no dog in this hunt. That being said, I’m a little surprised that glee was surprised that AI was able to suss out who he is. Chess at the National/Intl. ranked level is pretty rarified; it doesn’t take too many tidbits of information, along with things like age, for someone to put two and two together with some Google searching. AI would have an easier time with it.
I asked Claude (Sonnet, 4.6, adaptive) “I am “JohnT” at boards.straightdope.com. What can you tell me about myself just from my posting history? Go ahead and spend some tokens on this, thanks.”
It was about 90-95% correct, even though it misidentified me as Stranger on a Train in one minor instance. Here’s the full analysis:
I then corrected it, saying "That’s pretty good, except it wasn’t me who opened the “What’s your problem, Chronos?” thread… that was a poster named “Stranger on a Train”. I also don’t talk about orbital mechanics or ISS physics.
And, in case you care, the one thread I seem to be most remembered for is my Medicare thread. I still get asked questions about it 6 years later:
America's Elder Crisis, Medicare Edition"
To which Claude responded:
TLDR version: Yes, this is a thing.
I don’t think this is really about AI. I saw the post in question, and being interested in the topic, followed a link. It was pretty obvious there which person was the poster in question, although the name meant nothing to me and I couldn’t pick it out of a lineup now.
But this wasn’t a case of AI being able to match up a bunch of tidbits scattered across a poster’s history. It’s the same thing that has always been true – it’s probably easier for people to figure out who we are than we sometimes think.
Further to that, in a discussion some time ago about a major data breach, someone – I think it was @LSLGuy – correctly pointed out that being overly obsessed about privacy is essentially futile. I don’t remember the exact words, but the essence of it was that for most of us, there’s probably a lot more personal information floating around out there than we realize, likely including things like our SSN and birth date.
One should take reasonable precautions against identity theft, but beyond that excessive fretting is pointless. Here in Canada, the federal government urges citizens to keep their Social Insurance Number (SIN) confidential, but won’t issue a new one even if the old one was compromised, unless you provide evidence of actual identity theft rather than just the potential for it. Likewise, the birth date of anyone important enough to have, say, a Wikipedia entry will usually be public information, even though it’s often part of identity verification with many organizations.
I believe he’s given even more specific information than that. I just don’t know the specific skills accessible to an LLM to use that info.
AFAChatGPTK (new acronym!) Kent_Clark is likely either a lawyer or public official with a strong background in science who lives in some indeterminate part of the U.S.
Actually, I do live in a definite part of the U.S. As for the rest of it, let’s just say my secret identity is working way better than I expected.
What, is the Daily Planet behind some sort of super-secure paywall?