Are questions about one's own health private? Or can OpenAI and Anthropic sell the fact that you, personally, asked XYZ question to insurance companies? Employers? If you ask something about you're feeling after you took some cocaine going to be forwarded to LEO's and/or insurance companies and employers? If you express that you distrust capitalism (or hate DJT), will it be forwarded to the FBI, so they can monitor you for 'seditious' thoughts? Et Cetera.

The entire question of what's protected WRT what someone says to/asks an AI seems to still be very much an unanswered question, and the topic is potentially VERY important.

Why in the world would you expect the tech bros behind that NOT to use every bit of data you give them in every way that's convenient and/or profitable for them?

I never say anything to an AI that could positively peg me as a hater of Trump, or capitalism, or give it any health secrets about myself or that would reflect badly on me as an employee. I absolutely ASSUME it's a giant surveillance apparatus and everything you say or ask could one day be used against me in some way.

I think its foolish to make any other presumption about them, because as I am pointing out once again here, the biggest problem with AI is the lack of proper regulations being laid down before it was widely rolled out. Which would've included, as you say, safeguards against them consuming non-public works in their training. Which YES, they could've done. Not necessarily 100% reliably (that much is true), but a hell of a lot more reliably than just doing nothing of the sort. And their products may have been less full-featured than what the Sam Altman's of the world felt was "critical", but TOO FREAKING BAD.

was releasing badly flawed tech that couldn't be stopped from hallucinating and could make mistakes at any time in almost any way - and then shifting responsibity for the errors to users. The next biggest was hyping badly flawed technology. The industry is built on theft and fraud.

1. The "Learning vs. Copying" Distinction
The most common misconception is that AI is a giant database of images or text that it "remixes." It isn't.The Argument: When an AI is trained, it doesn't "save" the images or text. It converts them into mathematical weights (numerical values representing patterns). The Analogy: If you read a thousand mystery novels and learn that "the butler did it" is a common trope, you haven't stolen those books; you've learned the statistical probability of a plot point.The "K.O." Line: "The model doesn't contain a single pixel or word of the original data. It contains the logic of how those things are formed. If learning from a public work is 'theft,' then every art student in a museum with a sketchbook is a shoplifter."

2. Transformative Use (The Legal Shield)
In the U.S., the Fair Use doctrine is the biggest hurdle for the "theft" argument.The Argument: For something to be a copyright violation, it usually has to act as a market substitute (i.e., people buy the copy instead of the original). The Fact: In recent 2025/2026 rulings (like Kadrey v. Meta), courts have found that training is "spectacularly transformative." The purpose of the data isn't to be "re-stated" but to be used as a "biological" input to create something entirely new. The "K.O." Line: "Copyright protects the expression of an idea, not the facts of its existence. AI analyzes the data for functional patterns, not to reproduce the art. Use is only 'theft' if the output is a 1:1 clone, which is already illegal under existing laws."

3. The "Style is Not Copyrightable" Reality
Many people are angry because AI can mimic a specific artist's "vibe."The Argument: Legally, you cannot copyright a style, a technique, or a genre. You can only copyright a specific, finished work. The Reality: If a human paints "in the style of Van Gogh," we call it an "homage." If an AI does it, people call it "theft." This is a double standard.The "K.O." Line: "If you could copyright a 'style,' then every Impressionist after Monet would owe his estate royalties. You’re mad at the efficiency of the tool, not the legality of the process."

especially with that stupid bot offering what are laughably described as "K.O. lines." Did you suggest that wording?

1. AI does save text, as was proven conclusively by research at Stanford that got plenty of media attention late last year, as large chunks of text and nearly complete books were extracted from various genAI models. Your chatbot not mentioning that is evidence of pro-AI company propaganda.

2. The court battle is far from over, with no consensus in the rulings. Again, your bot spat out pro-AI propaganda. And even if there was any consensus in court that what the AI companies did was not IP theft, the legal battle to overturn that wrong decision will continue because there's ample evidence from what those in the AI industry have said themselves that they were aware it was theft.

3. I've never suggested style is copyrightable. I don't know of any AI opponent who ever did.

just another aspect of its sycophancy and it wanted you to think those would be knockout lines.

I was watching a video last night with an expert on guitar pedals, who's written books and has his own business, showing how unreliable ChatGPT was, how much it would get wrong, how it would hallucinate lengthy descriptions of nonexistent pedals if given a fake name of a pedal. He was wondering if there'd be any correct information left online for future generations if genAI bots keep running amok.

... and put all these data centers to good use.