Im not a lawyer but a lot of what "AI" does seems like it would be uh illegal or at least disastrously bad from a risk management POV.
-
Aahahaahah go fuck yourself google
Offering AI-powered research is "above all an expression of Google's business activities" and "at most a secondary expression of an interest in being able to freely express one's opinion and beliefs."
@jonny good thing that Germany doesn't have a "free speech" fetish.
One of the lessons we've learned from Fascim V 1.0
Edit: Alas these days I'm not sure any of those lessons have stuck.
-
@jonny if everyone knows not to believe it, then WHY IS IT THERE???
@sammy@cherrykitten.gay @jonny Remember the clause in Microsoft Copilot's EULA stating it is for entertainment purposes only? They are playing games, one being a rousing "catch me if you can" with the legal system.
-
(I'm reading the decision through crappy machine translate btw and hopefully its clear that's not an actual quote. I continue to not be a lawyer)
This is like real bad for AI.
Again, fresh in-post label for I am not a lawyer and I am reading this through shockingly bad PDF machine translation.
The judge is explicitly cutting down most of the legal defenses they use. They make a sharp cut between search and AI, saying search is indispensable, but AI is not, and defendants have not proven how being held liable for their output would compromise the ability to run a normal search engine. They make a similar hard cut between AI and autocomplete.
They go on at length about the nature of truth in utterances, and arrive at a conclusion that AI output has no protections for free expression because it isn't expressing shit - it has no beliefs, it is a commercial product only. There are two injunctions that are denied because they are not considered statements of fact, but the judge rules against google for all the ones that were, and concludes several are default considered false because the linked pages were irrelevant.
There is explicit differentiation from aggregating reviews and third party content, because the AI generated text and ideas that were not present in the input. There is also discussion about how there is no excuse for further violations just because its hard to control AI output, and contrasts this with how normal "report and takedown" protections work.
There is very little here that is specific to AI overviews in search, and almost all of the arguments apply to AI products in general. AI's only prayer of being remotely profitable must include advertising or shopping features, which means they absolutely must continue generating output that makes statements of fact about other companies. I know nothing about how German courts work, the article alludes to appeals, but if this ruling holds even just in Germany the ability to insure AI products disappears overnight and that makes the product nonviable.
Edit: Germans, German speakers, and I guess by some miracle if there is a German lawyer wandering by, please feel free to correct me and I'll edit the post
-
@jonny Your toots look good tho. Just had a look at it (I'm german) and it says what you tooted
Although the "Landgericht München" probably isnt that small...
@Numerfolt
Good to know! Thanks for double checking -
The judge is basically like "the basis for the entire argument that keeps search engines legal is that they are indispensable for understanding the flood of information online, and its impossible for search engines to evaluate all the context they index. If what you say about "everyone knows the AI overview is not to be trusted" is true then you are fucking cutting your own legs out from underneath you buddy boy. If the thing is not useful, your protections are even more gone"
@jonny And they argue that the AI overview shows content that's not in any of the sources, so it is definitely something they're responsible for
-
@jonny And they argue that the AI overview shows content that's not in any of the sources, so it is definitely something they're responsible for
@Numerfolt
I love the parts where (it seems like) the judge was like "we clicked the links and they had nothing to do with the claim and sometimes they say the opposite" -
The judge is basically like "the basis for the entire argument that keeps search engines legal is that they are indispensable for understanding the flood of information online, and its impossible for search engines to evaluate all the context they index. If what you say about "everyone knows the AI overview is not to be trusted" is true then you are fucking cutting your own legs out from underneath you buddy boy. If the thing is not useful, your protections are even more gone"
@jonny Another good one:
The judges argue that the AI overview isn't neccessary, because the search result lists are already fulfilling the function that gives the search engines their privileges.
Therefore these privileges don't extend to the AI overview -
This is like real bad for AI.
Again, fresh in-post label for I am not a lawyer and I am reading this through shockingly bad PDF machine translation.
The judge is explicitly cutting down most of the legal defenses they use. They make a sharp cut between search and AI, saying search is indispensable, but AI is not, and defendants have not proven how being held liable for their output would compromise the ability to run a normal search engine. They make a similar hard cut between AI and autocomplete.
They go on at length about the nature of truth in utterances, and arrive at a conclusion that AI output has no protections for free expression because it isn't expressing shit - it has no beliefs, it is a commercial product only. There are two injunctions that are denied because they are not considered statements of fact, but the judge rules against google for all the ones that were, and concludes several are default considered false because the linked pages were irrelevant.
There is explicit differentiation from aggregating reviews and third party content, because the AI generated text and ideas that were not present in the input. There is also discussion about how there is no excuse for further violations just because its hard to control AI output, and contrasts this with how normal "report and takedown" protections work.
There is very little here that is specific to AI overviews in search, and almost all of the arguments apply to AI products in general. AI's only prayer of being remotely profitable must include advertising or shopping features, which means they absolutely must continue generating output that makes statements of fact about other companies. I know nothing about how German courts work, the article alludes to appeals, but if this ruling holds even just in Germany the ability to insure AI products disappears overnight and that makes the product nonviable.
Edit: Germans, German speakers, and I guess by some miracle if there is a German lawyer wandering by, please feel free to correct me and I'll edit the post
companies tend not to sell products with infinite liability that can be trivially baited by anyone in a few, free tries.
-
@jonny Another good one:
The judges argue that the AI overview isn't neccessary, because the search result lists are already fulfilling the function that gives the search engines their privileges.
Therefore these privileges don't extend to the AI overview@Numerfolt
I loved that part, and how they didn't convince the judge that being liable for their product's output would hinder normal search, which is the only thing that matters -
The ole' Fox News defense, eh?
"We're _obviously_ entertainment. You can't take us seriously."
I am not entertained and _everyone_ obviously does not fucking _know_.
<sigh>
@401matthall @jonny
ha, i am also named Matt and was also reminded of the Fox News defense -
@hvdsomp
Its the only plausible defense and it sucks assI suspect that this means that Microsoft's entertainment-only fig leaf would not carry legal weight in Germany. If 'everyone knows this product is trash that shouldn't be used for anything important' doesn't absolve you of liability for one LLM use, it probably doesn't for any other.
-
This is like real bad for AI.
Again, fresh in-post label for I am not a lawyer and I am reading this through shockingly bad PDF machine translation.
The judge is explicitly cutting down most of the legal defenses they use. They make a sharp cut between search and AI, saying search is indispensable, but AI is not, and defendants have not proven how being held liable for their output would compromise the ability to run a normal search engine. They make a similar hard cut between AI and autocomplete.
They go on at length about the nature of truth in utterances, and arrive at a conclusion that AI output has no protections for free expression because it isn't expressing shit - it has no beliefs, it is a commercial product only. There are two injunctions that are denied because they are not considered statements of fact, but the judge rules against google for all the ones that were, and concludes several are default considered false because the linked pages were irrelevant.
There is explicit differentiation from aggregating reviews and third party content, because the AI generated text and ideas that were not present in the input. There is also discussion about how there is no excuse for further violations just because its hard to control AI output, and contrasts this with how normal "report and takedown" protections work.
There is very little here that is specific to AI overviews in search, and almost all of the arguments apply to AI products in general. AI's only prayer of being remotely profitable must include advertising or shopping features, which means they absolutely must continue generating output that makes statements of fact about other companies. I know nothing about how German courts work, the article alludes to appeals, but if this ruling holds even just in Germany the ability to insure AI products disappears overnight and that makes the product nonviable.
Edit: Germans, German speakers, and I guess by some miracle if there is a German lawyer wandering by, please feel free to correct me and I'll edit the post
@jonny Looks fine to me. Not a lawyer. However my lawyer friend says the judgement looks air tight and there's a snowball in hell chance that another court will disregard it.
-
This is like real bad for AI.
Again, fresh in-post label for I am not a lawyer and I am reading this through shockingly bad PDF machine translation.
The judge is explicitly cutting down most of the legal defenses they use. They make a sharp cut between search and AI, saying search is indispensable, but AI is not, and defendants have not proven how being held liable for their output would compromise the ability to run a normal search engine. They make a similar hard cut between AI and autocomplete.
They go on at length about the nature of truth in utterances, and arrive at a conclusion that AI output has no protections for free expression because it isn't expressing shit - it has no beliefs, it is a commercial product only. There are two injunctions that are denied because they are not considered statements of fact, but the judge rules against google for all the ones that were, and concludes several are default considered false because the linked pages were irrelevant.
There is explicit differentiation from aggregating reviews and third party content, because the AI generated text and ideas that were not present in the input. There is also discussion about how there is no excuse for further violations just because its hard to control AI output, and contrasts this with how normal "report and takedown" protections work.
There is very little here that is specific to AI overviews in search, and almost all of the arguments apply to AI products in general. AI's only prayer of being remotely profitable must include advertising or shopping features, which means they absolutely must continue generating output that makes statements of fact about other companies. I know nothing about how German courts work, the article alludes to appeals, but if this ruling holds even just in Germany the ability to insure AI products disappears overnight and that makes the product nonviable.
Edit: Germans, German speakers, and I guess by some miracle if there is a German lawyer wandering by, please feel free to correct me and I'll edit the post
@jonny *Bad for AI _in Germany_
-
This is like real bad for AI.
Again, fresh in-post label for I am not a lawyer and I am reading this through shockingly bad PDF machine translation.
The judge is explicitly cutting down most of the legal defenses they use. They make a sharp cut between search and AI, saying search is indispensable, but AI is not, and defendants have not proven how being held liable for their output would compromise the ability to run a normal search engine. They make a similar hard cut between AI and autocomplete.
They go on at length about the nature of truth in utterances, and arrive at a conclusion that AI output has no protections for free expression because it isn't expressing shit - it has no beliefs, it is a commercial product only. There are two injunctions that are denied because they are not considered statements of fact, but the judge rules against google for all the ones that were, and concludes several are default considered false because the linked pages were irrelevant.
There is explicit differentiation from aggregating reviews and third party content, because the AI generated text and ideas that were not present in the input. There is also discussion about how there is no excuse for further violations just because its hard to control AI output, and contrasts this with how normal "report and takedown" protections work.
There is very little here that is specific to AI overviews in search, and almost all of the arguments apply to AI products in general. AI's only prayer of being remotely profitable must include advertising or shopping features, which means they absolutely must continue generating output that makes statements of fact about other companies. I know nothing about how German courts work, the article alludes to appeals, but if this ruling holds even just in Germany the ability to insure AI products disappears overnight and that makes the product nonviable.
Edit: Germans, German speakers, and I guess by some miracle if there is a German lawyer wandering by, please feel free to correct me and I'll edit the post
@jonny For some strange reason, my general mood just improved dramatically
-
This is like real bad for AI.
Again, fresh in-post label for I am not a lawyer and I am reading this through shockingly bad PDF machine translation.
The judge is explicitly cutting down most of the legal defenses they use. They make a sharp cut between search and AI, saying search is indispensable, but AI is not, and defendants have not proven how being held liable for their output would compromise the ability to run a normal search engine. They make a similar hard cut between AI and autocomplete.
They go on at length about the nature of truth in utterances, and arrive at a conclusion that AI output has no protections for free expression because it isn't expressing shit - it has no beliefs, it is a commercial product only. There are two injunctions that are denied because they are not considered statements of fact, but the judge rules against google for all the ones that were, and concludes several are default considered false because the linked pages were irrelevant.
There is explicit differentiation from aggregating reviews and third party content, because the AI generated text and ideas that were not present in the input. There is also discussion about how there is no excuse for further violations just because its hard to control AI output, and contrasts this with how normal "report and takedown" protections work.
There is very little here that is specific to AI overviews in search, and almost all of the arguments apply to AI products in general. AI's only prayer of being remotely profitable must include advertising or shopping features, which means they absolutely must continue generating output that makes statements of fact about other companies. I know nothing about how German courts work, the article alludes to appeals, but if this ruling holds even just in Germany the ability to insure AI products disappears overnight and that makes the product nonviable.
Edit: Germans, German speakers, and I guess by some miracle if there is a German lawyer wandering by, please feel free to correct me and I'll edit the post
@jonny hell yeah!
-
@jonny *Bad for AI _in Germany_
@wronglang
Its hard to be exposed to liability constantly to everyone in the entire EU (they make an international claim of jurisdiction) -
@jonny Your toots look good tho. Just had a look at it (I'm german) and it says what you tooted
Although the "Landgericht München" probably isnt that small...
@Numerfolt @jonny does Germany have an "appeal to a higher court" kind of mechanism that would apply here?
-
Its awesome how Google's defense here is "everyone knows not to believe the AI overview"
@jonny that was the same reason Fox “News” got away with defamation. Nobody with any sense is supposed to believe that they’re actually reporting real news.
Unfortunately people do.
-
@wronglang
Its hard to be exposed to liability constantly to everyone in the entire EU (they make an international claim of jurisdiction)@jonny well that's something
-
Im not a lawyer but a lot of what "AI" does seems like it would be uh illegal or at least disastrously bad from a risk management POV. tight race between financing or insurance for what will precipitate the pop
https://the-decoder.com/landmark-german-ruling-declares-googles-ai-overviews-are-googles-own-words-and-makes-it-liable-for-false-answers/@jonny I like that the reporter avoided the word "hallucinate".
