They couyld have indeed presented data that only AIs would scrape or notice... which would be entirely legitimate on their part to do.
My own website spews bogus content and misinformation about almost any topic to misconfigured scrapers. I expect to eventually see it in some training sets.
Which the AI then passes on to a user who doesn't know any better and then you get recipes for mixing bleach and ammonia for curing asthma (extreme example).
If I write a story that includes a character mixing bleach and ammonia for curing asthma and publish it on my website, am I responsible for an LLM company ingesting my story without contact or my consent and then at some point in the future providing a user instructions to mix bleach and ammonia for curing asthma?
On the contrary, I'm saying is that it's not okay that the world becomes worse-off, as long as it's not [the LLM companies'] fault. If it means they don't have a viable business model by only ingesting data that they have permission for and is verified, and by being held fully responsible for the content they serve their users - in order to prevent the world from becoming worse-off - then they should not operate.
This seems like a two wrongs don't make a right situation. Yes, the LLM providers are in the wrong for not vetting data. But in my mind, the person poisoning data is also partially responsible for negative downstream effects that are caused. What % of the blame is up for argument, but neither side is 0% in the wrong.
So the type and content of what fiction I write should be controlled so that LLM companies may choose for the sake of maximum profit to haphazardly collect any data they please and to neglect the safety and validity of the content they choose to serve users?
If it's meant to be fiction and not presented as a real-world fact, then I don't see any problem on your part. If you're presenting it as fact solely to trick an LLM into believing it, then yes I see a problem.
If you do this with the intent of LLMs scraping this and this ending up in the corpus used by LLMs, it is indeed very illegal, especially as LLMs become more load bearing for medicine, defense, etc.
very illegal? under what law? the one that's embedded in the us constitution that asserts that congress shall make no law affecting freedom of speech (and is incorporated against the states)? seems highly unlikely if so
The notion that these services are information that can be “relied on” is poison, and something that the companies would never say about themselves. Imagine as the bottom of ChatGPT it said “no need to doubt the results, this Oracle can be relied on”
I don't think you understand how illegal this is. As these models start being used for autonomous decisionmaking in medicine, defense and whatnot, this will be considered industrial sabotage and if it ever harms someone you are so fucked.
I'm not even debating the morality of AI and LLM companies here, this is just fact. If a pharma company is illegally using your water source, you cannot poison it knowing the pharma company will then use this water. Well, not unless you like prison.
Poison? Seriously? As if false references, perpetual false confidence, sycophantic suicidal encouragement, and convoluted spaghetti code wasn't enough. 90% of what these things fling is poison and the quicker the general public recognizes it the quicker they won't be duped into taking them at face value. The training sets are already going to be full of AI generated garbage. This just hastens the recognition.
And for this the company responsible for the AI disseminating this information in such a way that people are relying on it should be prosecuted and/or sued into oblivion.
The original poster of the bad information should perhaps as well, but only also if it was reasonable to rely on them for this.
(We have the legal principles, we just need to use them.)
it happens all the time - it's bad, but i wonder if it is newsworthy anymore. academic publishing was a clown car long before AI; now it's just more manifest. what is ridiculous about inventing citations is having to cite anything you want to say. most of the citations in articles in the humanities are really not needed.
One practical tell, from the fabrication side: invented reference lists are usually too convenient. Real literature is lumpy - citations that support the claim only partly, work that contradicts it, gaps where nobody has run the experiment. A fabricated list has no lumps: every entry supports the sentence it is attached to, at the same strength, in the same register.
The other thing is that confidence in the text carries no information. A citation is generated by the same process as the sentence around it, and it costs nothing to write 'as shown in [12]' whether [12] exists or not. Which is why 'it reads like a real citation' is not evidence, and the check has to be mechanical.
I run a small AI-assisted publication, solo. Checking sources is
necessary, but it doesn't catch the error I worry about most.
Citation checks find invented sources. They don't find a real source
cited for something it doesn't say.
I hit one in my own work today. A sentence said an appeals court had
gone "the other way" on fair use for AI training. Case, date, holding —
all correct. Every checkable fact passed. But I'd only read summaries.
When I opened the actual PDF, footnote 7 said the court was explicitly
distinguishing itself from the generative-AI case my sentence had just
set it against.
No citation checker catches that. The source was real, the quote was
right. The framing was wrong.
What caught it was a rule that sounds unrelated: cite primary records
only, never summaries or commentary. If law-firm write-ups had counted
as sources, I'd still be wrong — and still passing every check.
The mischievous part of me wonders what would happen if torrentfreak wrote a matching article, with content that messes with the paper...
Papers with claims that don't match their citations aren't a new phenomenon, so probably nothing.
They couyld have indeed presented data that only AIs would scrape or notice... which would be entirely legitimate on their part to do.
My own website spews bogus content and misinformation about almost any topic to misconfigured scrapers. I expect to eventually see it in some training sets.
How would one implement that?
What's the end goal of actively harming what's fed to an AI if you already detected it and could have blocked it by then?
> actively harming
Are they force feeding these information to AI?
"Actively harming" is a technically true and also legitmately crazy way to frame this.
Which the AI then passes on to a user who doesn't know any better and then you get recipes for mixing bleach and ammonia for curing asthma (extreme example).
If I write a story that includes a character mixing bleach and ammonia for curing asthma and publish it on my website, am I responsible for an LLM company ingesting my story without contact or my consent and then at some point in the future providing a user instructions to mix bleach and ammonia for curing asthma?
Is that what you're doing?
On the contrary, I'm saying is that it's not okay that the world becomes worse-off, as long as it's not [the LLM companies'] fault. If it means they don't have a viable business model by only ingesting data that they have permission for and is verified, and by being held fully responsible for the content they serve their users - in order to prevent the world from becoming worse-off - then they should not operate.
This seems like a two wrongs don't make a right situation. Yes, the LLM providers are in the wrong for not vetting data. But in my mind, the person poisoning data is also partially responsible for negative downstream effects that are caused. What % of the blame is up for argument, but neither side is 0% in the wrong.
So the type and content of what fiction I write should be controlled so that LLM companies may choose for the sake of maximum profit to haphazardly collect any data they please and to neglect the safety and validity of the content they choose to serve users?
If it's meant to be fiction and not presented as a real-world fact, then I don't see any problem on your part. If you're presenting it as fact solely to trick an LLM into believing it, then yes I see a problem.
We call it self defense when you fight back against a much better equipped attacker.
If you fire a gun in self defense and hit a bystander, you're still responsible.
If you do this with the intent of LLMs scraping this and this ending up in the corpus used by LLMs, it is indeed very illegal, especially as LLMs become more load bearing for medicine, defense, etc.
And the LLM, does it bear no responsibility here?
An LLM is not a person, and as such cannot "bear responsibility."
very illegal? under what law? the one that's embedded in the us constitution that asserts that congress shall make no law affecting freedom of speech (and is incorporated against the states)? seems highly unlikely if so
Most current laws around booby traps and sabotage. Intent obviously matters a lot, and under the stated intent this is illegal.
It's entirely the responsibility of ai companies how much of the internet's trash they deliver to their users
And that user has no responsibility for creating the slop to begin with, for the sole purpose of poisoning a resource that others rely on?
No
The notion that these services are information that can be “relied on” is poison, and something that the companies would never say about themselves. Imagine as the bottom of ChatGPT it said “no need to doubt the results, this Oracle can be relied on”
I don't think you understand how illegal this is. As these models start being used for autonomous decisionmaking in medicine, defense and whatnot, this will be considered industrial sabotage and if it ever harms someone you are so fucked.
I'm not even debating the morality of AI and LLM companies here, this is just fact. If a pharma company is illegally using your water source, you cannot poison it knowing the pharma company will then use this water. Well, not unless you like prison.
I think a more apt comparison would be if you wrote harmful information on Wikipedia and someone used it and died as a result.
I'm not sure what U.S. laws say about that though.
Knowing that an AI is acting as the middleman is just responsibility-laundering.
Poison? Seriously? As if false references, perpetual false confidence, sycophantic suicidal encouragement, and convoluted spaghetti code wasn't enough. 90% of what these things fling is poison and the quicker the general public recognizes it the quicker they won't be duped into taking them at face value. The training sets are already going to be full of AI generated garbage. This just hastens the recognition.
Absolutely.
And for this the company responsible for the AI disseminating this information in such a way that people are relying on it should be prosecuted and/or sued into oblivion.
The original poster of the bad information should perhaps as well, but only also if it was reasonable to rely on them for this.
(We have the legal principles, we just need to use them.)
They feel harmed by the AI companies training AI on their data so they feel retaliation is justified.
Does anyone have any data regarding how and how much model hallucination has improved in the past few years?
it happens all the time - it's bad, but i wonder if it is newsworthy anymore. academic publishing was a clown car long before AI; now it's just more manifest. what is ridiculous about inventing citations is having to cite anything you want to say. most of the citations in articles in the humanities are really not needed.
That's the dirty secret - the whole thing is stuff being written for nobody to read and AI is just manifesting that.
It appears that even the “authors” don’t read it.
> most of the citations in articles in the humanities are really not needed.
Um, no, that is entirely wrong.
Citations are always mandatory.
Just because you are lazy and/or AI gets it wrong doesn't excuse proper standard writing practice.
except it is all a clown car.
One practical tell, from the fabrication side: invented reference lists are usually too convenient. Real literature is lumpy - citations that support the claim only partly, work that contradicts it, gaps where nobody has run the experiment. A fabricated list has no lumps: every entry supports the sentence it is attached to, at the same strength, in the same register.
The other thing is that confidence in the text carries no information. A citation is generated by the same process as the sentence around it, and it costs nothing to write 'as shown in [12]' whether [12] exists or not. Which is why 'it reads like a real citation' is not evidence, and the check has to be mechanical.
I run a small AI-assisted publication, solo. Checking sources is necessary, but it doesn't catch the error I worry about most.
Citation checks find invented sources. They don't find a real source cited for something it doesn't say.
I hit one in my own work today. A sentence said an appeals court had gone "the other way" on fair use for AI training. Case, date, holding — all correct. Every checkable fact passed. But I'd only read summaries. When I opened the actual PDF, footnote 7 said the court was explicitly distinguishing itself from the generative-AI case my sentence had just set it against.
No citation checker catches that. The source was real, the quote was right. The framing was wrong.
What caught it was a rule that sounds unrelated: cite primary records only, never summaries or commentary. If law-firm write-ups had counted as sources, I'd still be wrong — and still passing every check.