It’s just a ruling from a lower court, but it sets the stage for how European courts are going to deal with the question of who is liable for whatever slop “AI” generates.
The Regional Court of Munich hit Google with a temporary injunction barring the company from spreading false claims about two Munich-based publishers through its AI-generated search overviews (case no. 26 O 869/26). The court classified Google as a direct infringer because the “AI overview” is its own content, not just a list of search results.
Google’s AI overviews had falsely tied two publishing companies to scams, subscription traps, and shady business practices for certain search queries. According to the court, the AI mixed up information about other, genuinely sketchy companies with the plaintiffs and drew connections that didn’t appear in any of the linked sources. The publishers sent Google a cease-and-desist letter, but Google didn’t respond appropriately.
↫ Matthias Bastian at The Decoder
Google tried to argue it doesn’t carry any responsibility or liability for whatever slop its “AI” generate, but the German court does not agree. According to the court, “AI” overviews are not the same as regular search results, because they rewrite findings and just make shit up, thereby making claims that are nowhere to be found in any search results (or in reality in general). Furthermore, the court states that Google develops the “AI”, it runs it, it offers it to users, and Google alone controls its output, and as such, Google is liable for whatever their “AI” produces.
Google also tried to argue that users know not to trust anything an “AI” produces, which is hilarious considering how hard Google is pushing these tools, but the courts state that the ability of users to do further research does not absolve Google of liability. In addition, the court made it very clear that free speech protections absolutely do not apply, because the “AI” expressions are coming from an algorithm, not a person, and are above all an expression of Google’s business activities”.
In other words, if an “AI” tool generates false accusations and misleading statements, the creator of said “AI” is liable. With this ruling in hand, countless other people have a stronger case to make whenever Google or any other company tries to absolve itself from liability from slop just because a pachinko machine generated it.
Excellent news, and the only fair outcome.
Excellent.
Another win for European bureaucracy, where they can now enjoy not being confronted with the latest technologies and their harms. Google will now have no choice but disable AI in their search results, and German citizens will have to manually scan and digest information, like 20th century!
There is nothing like artisanal, hand curated search of reading 4 pages of links. The smell of those pages, and being bombarded with ads.
Of course money is no matter. Even though European homegrown startup culture is bordering non existence, they amount of bridge toll, sorry fines for improper conduct, they collect from American companies more than make up for it. Yes, why would it matter if the entire European startup sector is X* billions, while they manage to extract Y* billions with this scheme
And they don’t need to worry about single source. Even if this compliance moat prevents any homegrown startup, Chinese frontier labs are also in full swing, and they produce excellent products. If things go sour with Americans, they can always source from a Chinese dependency. However, I’m not sure they are as malleable for paying those tolls, sorry again, the fines.
(I can’t look up X and Y again. It would require me to use an AI summary)
(And if it is not apparent, this is satire)
I’ve seen hundreds of truly insightful comments from you sukru (or şükrü?) but apparently satire isn’t among your strengths.
The definition of sovereignty is pretty much about the ability to apply tolls wherever you like. And if we assume the European people to be sovereign in their lands, they are free to tax whoever and whatever they like. On the other side of the coin, American tech companies never complain about paying half the rate of corporate tax because their headquarters are somehow located in Ireland.
And also there a lot of reports on how the tech sector in the EU is booming. See TLDR for instance https://www.youtube.com/watch?v=u_N3-FKbi5c
To be clear, I’m also on the same page with you, and swear loudly whenever I come across one of those shithead water bottle caps.
cevvalkoala,
When the reality is more amusing than satire, it becomes difficult to emulate.
I really think you’re missing the point. It’s really not about politics and certainly not about scientific/societal progress, but about a search engine trying to stay relevant by pitting its own (low quality) generated content against the search results that the user asked for. Google not being responsible (as much) for third party content makes sense, but extending that privilege to it’s own content… No, of course not.
Rrups,
I think there are some usual misconceptions here. Yes, it might look that way if secondary effects or how this actually works is not considered
But it is… EU putting up high tolls, means they inadvertently entrench the established big players. A $1 billion fine could be “cost of business” for a behemoth like Google, but absolutely devastating for a local startup.
And this shows. I was not joking when I said EU has make more money from US tech companies paying tolls, than homegrown organic ones. The entire startup sector is minuscule compared to what modern economies demand.
This only harms the local EU economy, since American companies can continue to extract value from their market, and also extract talent who would not want to be shackled by local politics.
The LLMs, by definition are summarization engines. Hence the “content” is still not Googles.
Ironically, Google hosted their own content for over a decade, in the form of Knowledge Graph. The “info boxes” you see on the right are not entirely organic (unlike LLM summaries), but comes from an internal hand curated database.
https://en.wikipedia.org/wiki/Knowledge_Graph_(Google)
What is the benefit of the AI summary?
That is a long discussion, but basically saves time, while still preserving links to the original content.
“But it is… EU putting up high tolls, means they inadvertently entrench the established big players. A $1 billion fine could be “cost of business” for a behemoth like Google, but absolutely devastating for a local startup.”
No.
Normaly these fines scale with revenue.
smashIt,
Again, this feeds back to what I wrote.
In an ideal world, yes. In real life, there are secondary and tertiary effects.
I would go into details on how EU tech sector is far cry from financially sustaining itself. But not need. The very sentence explains it… in a sad way
There are no large EU tech companies to fine in the first place.
Once tech behemoth Europe is entirely dependent on American and Chinese suppliers.
The question is:
Why?
This article is not about fines, but about a court ruling on what Google is responsible for and forcing Google to act in a responsible way.
I don’t agree that LLM are summarization engines by definition. LLMs are just really good at putting related (or just adjacent) words together.
Sure, a LLM can easily discard words or parts of sentences that are very generic and therefore likely not that important to be part of a summary. But although all those statistics can help with understanding text, it’s not the same as actually understanding text.
But even if they are, Google is not just using some of the shelf model that they can’t steer.
If Google wants to provide the answers, it’s liable for those answers. That’s the cost of stealing millions of documents and trying to turn a once-reliable “Search Engine” into an unreliable “Knowledge Engine.”
This comes from the same company that made an AI which called white people dangerous, refused to generate caucasians in images, and suggested you might be racist for asking. Google is not a responsible company given the power they have, and without their absolute stranglehold on digital infrastructure nobody would tolerate them.
Kver,
I can see the reason for making google liable for content on google’s website. It’s pretty strait forward why the courts should hold google accountable for content google created. However it doesn’t make as much sense to say it was “stolen”. Websites are free to block google and they’ve always been able to do it using webmaster tools, robots.txt, and other mechanisms. Most websites, even those that block others are actually allowing google’s bots.
Something like that should include a link so everyone’s on the same page as to what you are referring to. But I’d say there are conflicting goals between letting the AI produce what it does naturally versus imposing artificially censorship requirements. What you are describing sounds like one of those rules that’s been added to steer LLM output in a certain direction. Government regulation tends to illicit artificial rules on top of the AI, but then users criticizing AI for the effects of those rules is kind of ironic since humans did that, not the AI.
Alfman,
The same discussion happened when Google News was forced to pay for content they linked. I think it was France, but it could also be Germany again.
The government passed a rule where companies could ask for money to be featured in Google results. So Google naturally chose smaller, alternative news sources that offered their content for free.
The result?
Those big publishers begged to be back on the listings. They basically wanted to be paid for free advertisement, and got what they asked for.
The mutual agreement benefits all parties.
And I’m sure new negotiations will take place for being featured in these new summaries.
sukru,
The flip side of that coin is that google wants to use content for free…and largely gets away with it displaying google ads using other people’s scrapped content. This is why the lawsuit came about. Ideally all parties could negotiate a fair deal, but it’s just not realistic when one party is so much more powerful than the other party.
Alfman,
This is too close to home.
I cannot vouch for recent Google. But back when I was in Google News, we were not a profit making division. In fact, I don’t think we directly made any substantial money at all. We just served News all around the web, respected publisher boundaries, only featured a simple title and a small snippet
We did this while also respecting user’s intents and intelligence, by promoting sources all across the spectrum, so everyone would see all sides (unless they explicitly block certain publishers)
(There was some magazine subs in the app, but I don’t think that amounted to much)
However, those specific examples, a few countries wanted us to pay for the content we freely promote and help users discover them. So, not only the entire department was already working for no direct financial return, we would also essentially pay for bringing users to the publishers.
It never felt right.
sukru,
Google has vacuumed up the ad revenue and refuses to pay for journalism. Consequently the growth of google as an ad titan diverted ad dollars from funding traditional journalism. I’ll be honest, even to the extent you may be right, it’s hard for me to sympathize with google here; I wish more money had stayed in the journalism industry and less money had gone to google.
Of course some will say traditional journalism simply doesn’t deserve to survive if they can’t compete against the world’s googles. But I genuinely worry how the collapse of journalism, by way of taking away ad dollars and not paying journalists, threatens democracy itself. We’re creating an information divide and leaving people highly uninformed. While this may not have been google’s intention, regardless as advertisers flocked to google it inadvertently killed the business model that journalists were dependent on.
Alfman,
We literally had zero ads on Google News. And also News section in Search.
Where did you get the idea we “vacuumed” revenue? My colleagues were were flying to publishers and directly talking with them to improve their websites and hence their revenue. We even tried some “citizen journalism” project, but unfortunately it did not take off. And we even offered direct payment funds to keep publishers alive.
Where did you get that impression?
Exactly for this purpose we re-engineered Google News to focus on events, and provide a wide variety of views and sources on each news.
I think nobody in their right mind thinks that. We need journalism on the ground, not hundreds and hundreds of replicas of press releases and crappy online content.
The problem is… the content model itself became obsolete, and people are still trying to find ways to keep them alive.
sukru,
.Google’s entire business model came at the expense of traditional media that funded journalism…it seems very naive to suggest google’s growth did not come at great expense for journalists.
Ok, but google has not replaced funding for journalism in proportion to the advertising revenue they took away. I’m not asserting what you say is untrue, however any benefits you want to cite still have to be weighed against this backdrop of massive losses to journalistic funding.
sukru,
I guess I should clarify that when I say google, I mean the company at large and not pointing the finger at your specific division. I believe you if you say the people you were working with wanted to help journalists even if the google monopoly was hurting them economically.
Alfman,
No worries. As I said, I cannot vouch for what is happening today.
Hopefully that “do no evil” spirit will one day come back.
I really don’t see how Google has more to gain than news companies from being featured in search results or on Google News. And I can’t help but view a world in which you’re forced to pay in order to be able to link to news articles as a little dystopian.
The problem with news not being profitable anymore is real and very saddening. But Google links to new sites would be the last thing I’d attribute it to. Unfortunately people’s attention has been grabbed for some time now by social media, with worse polarization than traditional news. This is a real effect that has been studied and is being abused, as evidenced by the amount of money going into it. If anything, algorithms for Google social sites seem to me the least polarizing (though I don’t have any data here, feel free to correct me if I’m wrong).
cheemosabe,
Ok, but I did not attribute it to that. Admittedly this may have only clear to sukru, but I was not blaming his department for working with news sites. Rather it’s the broader impacts google had on redirecting the flow of advertising dollars.
Oh no doubt. It’s a one-two punch between the loss of advertising dollars for journalism and the increasing emphasis of social media replacing news departments.
Good. I use LLMs a lot these days and I know that especially the newer flagship models can be genuinely useful and surprisingly sophisticated, if they are given some time to “think” and verify their “thoughts” before answering.
However, the zero-effort “you have 0.5 seconds to write something” flash model they put on the Google front page is just horrible and it would be a benefit to everyone if it’s just simply taken away. Its output is confident looking but usually range from misleading to plain wrong. It does more harm than good.
Yes, exactly, the fast model they use on search results isn’t ideal. Not only that but it’s a sort of bait-and-switch in which they use AI generated content on a page where people expect to find more accurate data. A separate page where it’s clear the content is summarized by AI would be more honest.
But I worry that this sets a somewhat dangerous precedent. This kind of content, with its degree of reliability, should be allowed to be served in some way, for people who find it useful.
Totally hilarious! These companies want all of the money but none of the responsibility. Very capitalist. Very American.
What does Google’s AI have to do with pachinko machines?