Google made their AI overviews a part of their search engine product specifically so people wouldn't have to actually click the links. They went from "here's a list of places what you're looking for might be" to "this is the information you're looking for."
And when you do that, you're taking responsibility for it. If google wants to play the editorial-summation game, part of that is going to be taking lumps when those editorials are slanderous or otherwise cause harm.
Really, that's the sane way to handle AI-related issues. Policing what can be prompted or generated has strong free speech issues, and while I recognize some limits are pragmatically required, anything that comes out of a generator is functionally a private sketch, doodle, or note until someone decides to publish it. Once published, any responsibility for the nature of the material falls upon the person who published it, and should be treated the same as any other similar piece of analogue material by the law.
Changing general topics for the comparison: If someone uses Midjourney to infringe a Disney copyright, or Suno an RCI one, or whatever, and they don't publish it, that's no different than someone doing that in a sketchbook or a private recording.
If I publish a cover of a song my grandparents danced to at their wedding and put it up for others to listen on youtube without paying the Music Industry vig, I'm the one they'd go after, not Ableton Live for making the software I used to do it, or Fender for making the guitar I used.
As long as there's a potential legal use of the generated novel data, and fair use provides plenty, addressing an issue with art at the art-tool level infringes the rights of the people who would use it that way. I tend to favor the expression right of the individual over the vague claims of financial harm by corporations, especially in these situations.
But the legal efforts are aimed where the money is, not where the fault is. Which is how the media giants like it, because if they have to go after individual people, every instance is a chance for their rent seeking to become a public relations nightmare like in the napster days.
Swinging back to Google, under these principles, google has very clearly taken responsibility for the nature of their AI summations by taking on the role of publisher. If you publish something slanderous or dangerously incorrect without human oversight, that doesn't get you off the hook. Google decided to become, in essence, a news source, and that's cool. It can do that, but it puts itself under the same rules as other news sources when it does so.
Not that Google wasn't acting as a publisher before, but Google's other ways of manipulating your search results for their own ends, like secretly adding "for sale" to searches behind the scenes, it's neutering of its own image search features and google cache, and prioritizing advertising partner and sales results over informative ones (all pre-AI lobotomies of a once very useful utility) were sadly too vague for courts to latch onto. The situation here is less obscured by technological plausible deniability and action was thus possible.