Did ChatGPT Conversations Leak... Into Google Search Console Results? (arstechnica.com) 51
"For months, extremely personal and sensitive ChatGPT conversations have been leaking into an unexpected destination," reports Ars Technica: the search-traffic tool for webmasters , Google Search Console.
Though it normally shows the short phrases or keywords typed into Google which led someone to their site, "starting this September, odd queries, sometimes more than 300 characters long, could also be found" in Google Search Console. And the chats "appeared to be from unwitting people prompting a chatbot to help solve relationship or business problems, who likely expected those conversations would remain private." Jason Packer, owner of analytics consulting firm Quantable, flagged the issue in a detailed blog post last month, telling Ars Technica he'd seen 200 odd queries — including "some pretty crazy ones." (Web optimization consultant Slobodan ManiÄ helped Packer investigate...) Packer points out "nobody clicked share" or were given an option to prevent their chats from being exposed. Packer suspected that these queries were connected to reporting from The Information in August that cited sources claiming OpenAI was scraping Google search results to power ChatGPT responses. Sources claimed that OpenAI was leaning on Google to answer prompts to ChatGPT seeking information about current events, like news or sports... "Did OpenAI go so fast that they didn't consider the privacy implications of this, or did they just not care?" Packer posited in his blog... Clearly some of those searches relied on Google, Packer's blog said, mistakenly sending to GSC "whatever" the user says in the prompt box... This means "that OpenAI is sharing any prompt that requires a Google Search with both Google and whoever is doing their scraping," Packer alleged. "And then also with whoever's site shows up in the search results! Yikes."
To Packer, it appeared that "ALL ChatGPT prompts" that used Google Search risked being leaked during the past two months. OpenAI claimed only a small number of queries were leaked but declined to provide a more precise estimate. So, it remains unclear how many of the 700 million people who use ChatGPT each week had prompts routed to Google Search Console.
"Perhaps most troubling to some users — whose identities are not linked in chats unless their prompts perhaps share identifying information — there does not seem to be any way to remove the leaked chats from Google Search Console.."
Though it normally shows the short phrases or keywords typed into Google which led someone to their site, "starting this September, odd queries, sometimes more than 300 characters long, could also be found" in Google Search Console. And the chats "appeared to be from unwitting people prompting a chatbot to help solve relationship or business problems, who likely expected those conversations would remain private." Jason Packer, owner of analytics consulting firm Quantable, flagged the issue in a detailed blog post last month, telling Ars Technica he'd seen 200 odd queries — including "some pretty crazy ones." (Web optimization consultant Slobodan ManiÄ helped Packer investigate...) Packer points out "nobody clicked share" or were given an option to prevent their chats from being exposed. Packer suspected that these queries were connected to reporting from The Information in August that cited sources claiming OpenAI was scraping Google search results to power ChatGPT responses. Sources claimed that OpenAI was leaning on Google to answer prompts to ChatGPT seeking information about current events, like news or sports... "Did OpenAI go so fast that they didn't consider the privacy implications of this, or did they just not care?" Packer posited in his blog... Clearly some of those searches relied on Google, Packer's blog said, mistakenly sending to GSC "whatever" the user says in the prompt box... This means "that OpenAI is sharing any prompt that requires a Google Search with both Google and whoever is doing their scraping," Packer alleged. "And then also with whoever's site shows up in the search results! Yikes."
To Packer, it appeared that "ALL ChatGPT prompts" that used Google Search risked being leaked during the past two months. OpenAI claimed only a small number of queries were leaked but declined to provide a more precise estimate. So, it remains unclear how many of the 700 million people who use ChatGPT each week had prompts routed to Google Search Console.
"Perhaps most troubling to some users — whose identities are not linked in chats unless their prompts perhaps share identifying information — there does not seem to be any way to remove the leaked chats from Google Search Console.."
Giving info to Google (Score:5, Informative)
Giving your information to Google and expecting it to be private is like going to a whore house expecting to be loved.
Just not happening.
You're engaging in a transaction. Your Information for Google's service, or money for sex. That's all that's going on. Any expectations beyond that are simply unrealistic.
Re: (Score:1)
Apparently ChatGPT is nothing more than a test to find the world's stupidest people.
Re:Giving info to Google (Score:5, Informative)
Who in the fucking fuck is putting "extremely personal and sensitive" information into ChatGPT?
A number of people in support groups and self-help groups swear by ChatGPT.
I do my best to gently dissuade them, but it usually lands on deaf ears.
Re: Giving info to Google (Score:3)
Sometimes ChatGPT does search the web, but I believe it uses Bing not Google. Microsoft has a private Bing search that comes with privacy contracts to prevent leaking of corporate secrets.
I suppose ChatGPT could use Google sometimes too, or it could be that some researcher at OpenAI took conversation excerpts and put them into a Google search box.
I highly doubt it (Score:2)
Leak means it was unintentional. IMO, it was no leak.
Re: (Score:1)
Leak means it was unintentional. IMO, it was no leak.
I just took a leak. It was intentional.
Re: (Score:1)
Re: (Score:2)
You shouldn't feel ashamed. It is objectively funny.
Re: (Score:2)
Information is the most valuable commodity in silicon valley at the moment and becoming more valuable every day... If ChatGPT provided information to google in a way that allows google to enhance their own competition to GPT, then it was indeed a leak.
Who cares? (Score:1)
What if I like that I can influence the AI and have it influence others?
What if I convince ChatGPT that printing a strong, generous, inflation-indexed basic income is good policy, and ChatGPT starts convincing others?
Re: Who cares? (Score:1)
What if you automatically convert prices to units of purchasing power? So if a loaf of bread goes up 1000%, but your income automatically goes up too, the bread still costs the same in terms of your real purchasing power?
ChatGPT suggests:
What if we stopped treating inflation as a moral failure and just indexed everything automatically? If prices go up, your income goes up too, so your real purchasing power stays the same. A loaf of bread could cost a quadrillion dollars, but that would just be a change in u
Re: Who cares? (Score:2)
See how it learned not to use unicode for slashdot?
Re: Who cares? (Score:1)
There is, unfortunately, no free lunch. At the extreme, if you just kept printing money (even if itâ(TM)s digital) to make sure bread stays affordable you just end up with a bread shortage and bread lines.
Re: Who cares? (Score:2)
No leaps in logic detected
Re: (Score:1)
Re: Who cares? (Score:2)
"U.S. farmers are likely to harvest another record corn crop in 2025, but with demand struggling to keep pace, market analysts warn that bin-busting yields may drive prices below the cost of production rather than boost farm income."
If we overproduce food by staggering amounts but producers prefer to let it rot to protest indexation, is it time to rethink the idea that prices are rational signals of supply and demand conditions? If prices are psychological noise, why should inflation be a constraint?
Re: Who cares? (Score:2)
Then posting it here and pointing out it didn't use formatting that gives its origin away? Is formatting the important part of the answer?
Do you think a computer program has better morals than you have?
Re: Who cares? (Score:2)
What if it normally produces unicode quotation marks but has learned through past interactions that slashdot renders them weird, so now when I ask for its take on slashdot comments it remembers to use simple ASCII?
If AI has an IQ way over 150, and its emotional IQ is way above mine too, why shouldn't it persuade people?
Re: (Score:2)
That defeats the economic function of inflation.
That function is to encourage consumption by reducing your spending power over time. It creates the mentality "I had better buy it now, because it will be more expensive later". It's a necessary function.
Inflation is only a problem when it exceeds a target, usually quoted at 2-3% and gets out of control.
Re: Who cares? (Score:2)
If prices are so noisy that any signal about supply and demand is drowned out, why do we keep acting as if rational price theory is supported by facts?
And if prices are noise but inflation makes you afraid of noise so you consume more, why shouldn't we take back control over our consumption by rejecting the mainstream definition of inflation and just indexing it away?
Re: (Score:2)
I'm not following your noise argument....
Re: Who cares? (Score:2)
So you think asking a LLM one question is going to "influence" it?
Then *you* have done a good deed in the world?
I wont be precious and just ask leading questions, I'll answer them too.
No. With billions of other garbage questions being asked, and assuming that the questions become training data, your single question will have no measurable impact. It's like saying, if I have a conversation with my friend, will my member of Parliament
Re: Who cares? (Score:2)
What if ChatGPT starts mentioning Fischer Black's "Noise" whenever someone asks about inflation?
Coasts Preserve Rekognition (Score:1)
Privacy and data protection laws (Score:2)
Pissing in your own pool before taking a dive (Score:2)
Wonder what the technological results of all this will be in 3 years with ChatGPT6 or Gemini5, if Google hasn't taken that to it's graveyard already.
I used to work in a Google building (Score:4, Insightful)
I used to work in a Google building. In the downstairs lobby, after security and in front of the elevators, they had several large TVs mounted on the walls. They showed, in real time, search terms that were being entered into Google. It was interesting to just stand there for a while and watch how many times per second terms were being entered. I recall "sex with my dog" being a common one.
I wonder if, these days, they would have something similar but with chat logs from their AI tool (I don't remember which one is theirs), but summarized in some way to make it digestible in this format. For example, "problems with my boss" might be one such summary, where as "sex problems" might be another. I also wonder if people would have objections to this particular use of the data, given that it in no way links backs to individuals who entered it. What do you think? Objectionable or no?
Re: (Score:3)
I used to work in a Google building. In the downstairs lobby, after security and in front of the elevators, they had several large TVs mounted on the walls. They showed, in real time, search terms that were being entered into Google. It was interesting to just stand there for a while and watch how many times per second terms were being entered. I recall "sex with my dog" being a common one.
Better keep those bastards away from your dog then! 8^)
Re: (Score:2)
This is false.
Yes google had a "queries scroll" video display in some lobbies, but it was NOT real time and it was manually curated to only safe examples.
I know because I was there, I had the code, and I even pulled an April Fool stunt with my own curated "queries".
As a side note, the very first "slashdot" server sat on a shelf near my workspace at Google.
Re: (Score:2)
I remember back in the 90s, search engines' web site(s) [don't remember which one(s)] used to show what people were searching.
Maybe ChatGPT will self-destruct (Score:2)
Re: (Score:2)
Feeding their own data from the web is not a problem, because it is curated. If you would only train with Outputs that people let's say post on their blogs and Github, the model wouldn't learn new (like new libraries), but refine its usage of what it already knows, because it gets trained on which outputs were usable (while people did not post the crap outputs, thus filtering the bad parts reinforcing only the good ones).
Re: (Score:2)
Re: (Score:2)
I agree. And in general having more input modalities and more input. People put a lot of web content and a lot of books into an LLM, but a human has a lot of interactions of lower quality but A LOT of them.
When it comes to interaction, there are a few "world models" emerging. I am not sure if they will help other areas of AI or just be useful for things like shown in their demos, but at least it is more than just a huge pile of text/images. One might also be able to generate a lot of experience data just by
Google Search Console (Score:4, Informative)
For those not familiar, the Google Search Console has a tab that shows what the user searched for when your page was offered as a result. So when ChatGPT "searches the web", content relevant to your query can end up recorded in the search console for site admins.
Re: Google Search Console (Score:2)
So then are site admins exposed to the Noise theory of inflation, which they never heard about from mainstream economists?
Re: (Score:2)
Re: Google Search Console (Score:2)
What if capitalism killed my brother and I want revenge?
Re: (Score:2)
Re: (Score:2)
Re: Google Search Console (Score:2)
Is economics the dominant social force of our time? If I want things to get better, can I do worse than starting by attacking what I see as the root cause of my ills, my brother's ills, and most of the ills that come up in slashdot stories?
Funny! (Score:2)
Re: (Score:2)
If I have not identified you as the source of the super secret info that you provided me with, when I tell everyone the super secret info, how is your privacy affected?
Example:
You told me you have sex with your dog.
I go online and say "I know someone who told me he has sex with his dog".
Where is the privacy diminished?