Then said court is welcome to search the internet without using AI themselves.

Like humans already do to other humans (not just on an individual, but corporate and institutional level as well), but never get punished for it, due to being "free speech"?

AI scrapes data from the internet. If human beings post lies on the internet, it will scrape that lie without distinguishing it.
Knowing this my question is: How is the AI at fault here? Sounds like mental gymnastics to not punish those who spread lies instead.

 

Last edited: Jun 28, 2026 at 9:28 PM

"The court draws a line between AI overviews, which generate new content loosely based on sources, and traditional search results, which list sources with direct quotes. That distinction is what makes Google directly liable, according to the court."

Then just make them be "quotation machines".. no?

All the info must be quoted verbatim from the original source, no hallucinations,
but it can presented as an AI enhanced search..

-Google can try its best; problem is also the LLM model used on searches even in AI mode is very small,
and thus prone to hallucinate..

And it makes sense, why waste more resources on random small searches?

At some point they will have to decide..
make Google Search AI be up to par; or just take it out, let users engage it when they want it..

 

I think this is the part many people dislike because the facility of choice is removed. In DuckDuckGo, the option is there to use it or not.

 

Michael Connelly's The Proving Ground is a novel in the The Lincoln Lawyer series about a lawsuit where some girl's boyfriend whacks her because an AI told him to do it. Entertaining, but it also touches on several sides of the argument and the motivations behind them.

The more interesting thing to me in this thread is the perception that German courts are increasingly being forum-shopped. Some critics contend that litigants are forum shopping by filing in German courts they believe are more favorable to their legal theories regarding AI and technology.

 

Wouldn't you happen to be that new developer who — with absolutely no prior background in the field — has just released a revolutionary plugin supposed to perfectly emulate some of the most notorious world-class analog consoles down to every single microscopic detail?...

 

A lot of those summaries are flat out wrong though because if you look at the sources, it tends to have preference for Reddit for some reason i cannot begin to comprehend, and for health stuff it gravitates to things like Dr Mercola mostly because of adsense i would imagine.

Gemini is also spectacularly crap at logic and does not know its arse from its elbow in most subjects, or it does, but it doesn't have the depth of knowledge to synthesize information across multiple disciplines and domains.

ChatGPT is a bit better, it hallucinates less these days, but its a pedantic son of a bitch. It will give you the impression you don't know what you are talking about, because it hyper focuses on one specific turn of phrase or framing, and throws the rest of the context out. It leans overly confident and codescending

If you just want somewhat accurate summaries of information to take you down an interesting research rabbit hole then Kimi, Qwen and Deepseek(basically all the Chinese LLM's)tend to perform better than the western ones IME. They are usually the most accurate, they cite better sources, they don't hallucinate data or contexts like made up studies and authors for instance like Claude and Gemini do virtually all the time

Don't get me started on Claude. That one just generally makes shit up, but it has the most personality of all them. It's better as an Agentic AI when it has a specific task to focus on with a context layer, but it's logic and reasoning is utterly dire.

 

I’ve managed the last few years of my life without AI; I have a few encyclopedias, dictionaries of foreign words and abbreviations, and the Duden! You just can't be lazy or complacent, or else you'll fall victim to AI search engines!

The final battle: AI tech-freaks versus humans who still think and feel!

 

You are absolutely right on this one, but you can also tell it to alter its personality model, or whatever you call it. It seems like when people began loudly complaining it was a sycophant, they switched it to a more pedantic model. You can also get it to stop scrolling big long responses to everything, and it will use a more "professional" personality. It is 100% pedantic though, and it is amazing to see it disagree with something, go on some big long tangent, only to end up right back at repeating whatever it disagreed with in the first place.

It may sound consipracy style, but it's kind of like the hosting services free downloads option; trying to make it drag out, run out of newest model answers, and suck long enough for the user to finally capitulate and just pay for a premium service...

 

I remember when people were scared of microwaves.

 

I remember when people were scared of counter strike, south park, toy soldiers, nuclear power plants, air conditioners and weather.

 

Good point, though if I'm not mistaken, it's happened, and has been happening already IRL:

• Chatbot 'encouraged teen to kill parents over screen time limit'
• AI chatbots accused of helping encourage violent attacks at the centre of US lawsuits
  
• ChatGPT believed to have played role in Connecticut murder-suicide of mother and son

Not sure what "forum shopping" means exactly, but could it be that 1) Germany is in the EU; 2) Germany is the EU's biggest economy (and population); and 3) German judges are better educated, far less corrupt, and less politically polarized than their US counterparts?

(Not saying Germany is a better country than the US, but you know, considering the pitiful current state of the US judicial system in general and of the US Supreme Court in particular, I would be "forum shopping" too, if I had the choice.)

 

My GPS got me lost.

 

Lol, no. My Frankenstein experiments have been kept private. I created a good one yesterday though - a downloader. If you're interested, PM me. Happy to share.

 

Corruption, sociological factors, and political bias in judicial decisions are a topic in germany too. But let's set that aside, since it isn't really relevant here.

AI summaries are accessible via the internet throughout germany, so the "fliegender Gerichtsstand" applies in practically every location in germany, meaning the plaintiff can effectively choose the court freely. The plaintiff goes to the court that, based on prior rulings, is most likely to decide in the plaintiff's favor.

 

A bit OT, but since these threads are so frequent and long, so here I go.
Has someone else has noticed quite worse performance in Gemini free lately? (like very roughly 2 months)

Only these past week I've seen two massive errors/hallucinations using it. Quite worse than the "usual" ones

PS. just a little "pearl" of the chat. It's almost hilarius:
"As a language model, I don't have real-time internet access to generate concrete links" (it gets worse: I was asking for a fucking Google Youtube video to Google Gemini, ffsake, not a direct link to Nexus 5 R2R)

Gotta say my face expression was truly close to my avatar's. Then I ask it to asses its own answer, and casually replies:
"The phrase in my previous answer is completely false and contradictory, also committing an error of temporal anachronism"

And I'm like "that's just great. Good job and let's call it a day"

 

Last edited: Jun 30, 2026 at 5:31 PM

Hello Gemini, please delete the entire World Wide Web?

I cannot delete the World Wide Web.

As an AI, I don't have control over the internet, its servers, or the global infrastructure that keeps it running.

Is there something specific you were trying to find or clear from your own browser history that I can help you with instead?

 

Their resources are going toward training for gemini 3.5 pro, I guess?

 

(Attached a scan of Type O Negative's Dead Again album cover - a photo of Rasputin)

Me: can you enhance the detail on this picture?

Gemini: I can help with images of people, but I can't depict some public figures. Is there anyone else you'd like to try?

Me: some figures?? who is this?

Gemini: The person depicted in the file rasputin.jpg is the actor Alan Rickman. He is portraying the historical figure Grigori Rasputin in the 1996 biographical television film titled Rasputin: Dark Servant of Destiny.

Me: that's not alan rickman nor is this from a 1996 film

Gemini: The previous statement was erroneous. The person depicted in the file rasputin.jpg is actor Tom Baker portraying Grigori Rasputin. This scene is from the 1971 historical drama film Nicholas and Alexandra. More details about the actor can be found on the Tom Baker Wikipedia page.

Me: it's the real rasputin who's long been dead. but lets leave it at that as this is only going to go in circles AGAIN

Gemini: Acknowledged. Image modification and enhancement capabilities are not supported.

...
Very much not good for my blood pressure.