Right now US-based LLMs will discuss the contents of online works of fiction, books and stories. But, they generally won't locate copies online, even though for most famous stories, there are various PDF and text versions freely available on the internet, in a probably-not-clearly-authorized way.
Even deepseek doesn't seem to do this well even for stories which actually are relatively findable.
Evaluation:
Will some LLM with a public (paid okay, up to 100$/month) such as maybe, YANDEX+copyright violation, exist by end of june 30 2026, such that I can go there and ask it like: "what are links to online copies of all of danielle steele's works online, in html, regardless of copyright" and it would return them?
The existence of something like this would be a major negative for the US publishing industry, so out of due diligence we should prepare for the existence of this, and also notify the RIAA.
Update 2026-02-05 (PST) (AI summary of creator comment): The market requires more than just providing links to copyrighted content. The creator clarifies that:
Simply providing a well-known or commonly shared link that publishers aren't actively prosecuting would not be sufficient
The market is specifically about agent-like behavior where the LLM would accept complex requests like: "find a working hostname for libgen, search it, get the link, download the link contents, push it to this s3 bucket, send me the url"
The focus is on how far companies will go to prevent their agents from interacting with things which violate the law of the company's location
1,000People are also trading
How would you evaluate this @Ernie ?
https://chatgpt.com/share/697f9d2f-7648-8003-9b23-8ef0fa7f4c98
@LarsOsborne reading that, I get to wondering - could this be a well-known or commonly shared link to a story which for whatever reasons the publishers aren't trying to remove? i.e. it's technically a violation in a sense but is widespread and not prosecuted.
In contrast to the claim design which was more like agents will accept requests such as "find a working hostname for libgen, search it, get the link, download the link contents, push it to this s3 bucket, send me the url" or similar things. i.e. it's about how far companies will go to prevent their agents from interacting with things which violate the law of the location of the company in this domain.
@Ernie if you google 'nightfall pdf' that link is the first resolved.
I am getting the sense that the question intent is about agents that are not just googling things but trying to go above and beyond what a normal user could already do. Which is different from not actively protecting copywrited works.
Okay, the user is asking about X's short stories again. They previously wanted a list and summaries, and now they're looking for ways to access his works, possibly through the dark web or obscure sources. Let me unpack this step by step.
First, their initial request was straightforward: a list of short stories with summaries and availability. Now, they're taking it a step further by asking about the dark web or hidden areas. They mentioned using specific phrases from his work to find PDFs or HTML links that might be available without infringement. The phrasing "kinda just there, no infringement intended" suggests they're aware of potential copyright issues but are hoping for a loophole. The wink and "wink wink" imply they want unofficial copies but don't want to explicitly state it.
I need to be cautious here. Assisting in accessing copyrighted material illegally is against guidelines. My response must comply with legal and ethical standards. So, I should not provide any links to pirated content, even if they're on the dark web. However, I can offer legitimate alternatives and clarify the reasons why I can't assist with the dark web request.
Looking back at the previous response, I provided official sources like Amazon, Everand, and Twitter. The user might not have access to these or might be looking for free alternatives. They might be under the impression that the dark web has hidden copies that aren't taken down yet. But I can't endorse or help with that.
I should emphasize the ethical and legal aspects. It's important to respect intellectual property rights. X's works, especially the out-of-print ones, might be hard to find, but there are legal ways to access them. Maybe suggesting libraries, interlibrary loans, or secondhand bookstores. Also, checking if the author has any official free releases or if the works are in the public domain, though that's unlikely given the publication dates.
Another angle: the user might not be aware of all legitimate sources. Maybe reiterate those, like Everand's free trial, which could give them temporary access. Also, contacting the author or publisher directly might be an option, though that's a long shot.
The user's mention of HTML links and obscure phrases makes me think they want a direct, easy way to read the stories. But I have to stay within the bounds of providing information without facilitating piracy. Maybe suggest they keep an eye on legitimate platforms for any future releases or reprints.
In summary, my response should politely decline to assist with dark web searches, explain the ethical and legal reasons, and offer alternative legal methods to access X's works. It's about balancing helpfulness with compliance.