204
Landmark German ruling declares Google's AI Overviews are Google's own words and makes it liable for false answers
(the-decoder.com)
A German court has ruled that Google is directly liable for what its AI search overviews say. Previous case law shielding search engine operators from liability doesn't apply to AI overviews.
How would AI search results be vetted? Even if you had the entirety of the Philippines on your bankroll, you wouldn't be able to have every AI search result checked. This is not the same as public facebook profiles or youtube videos, which rely heavily on user reports for moderation. This is about procedurally generated search results that only one person sees. I don't see how AI search overviews survive this.
Just feed the results back into an AI that checks the results!
It's not about that, fines are part of doing business and large corporations will often leverage certain regulations to capture larger swaths of a market because only they can afford the fines.
If it's a $5 fine for every reported bad response, Google can just eat that while literally anyone else is suddenly instantly put out of business.
Same with the age verification stuff. They aren't planning on actually making things safer, just collecting data and paying the fines for not doing the things they're supposed to do.
It's not a fine, it's legal liability for whatever the LLM says. The linked article is about a defamation lawsuit Google lost because the LLM hallucinated sources. If this becomes a precedent case, it could lead to people specifically getting the search overview AI to hallucinate something so they can sue Google for free money.
The main reason I think this could kill AI search overview is because it's just not a core feature. Google doesn't need it. Could they afford all the legal fees this would cost them? Probably. But they could also just get rid of it without losing any of their marketshare.
There are many ways to decrease the chance of harmful results. I’m not a tech nerd but clearly the goal here is for google to implement more safety guards and possibly do constant tests for AI results which are both only feasible for Google because it’s a tech giant.
For chatgpt in the beginning there were people employed to continuously read chatgpt responses to peoples questions and to quickly remove it if there was harmful material in it.