The New Law 2
Before we move on, we need to look at some misunderstandings of the law held by the general public, and the tricks and traps they lead to.
A.I. Furor
This German court action against Google seems to be a classic case of mountains out of molehills, but, as always, there's a hidden agenda at play.
And is everyone getting stupider in these comment sections? Both on Medium and YouTube, they're crowing that Google is being "brought to heel." Looks like they desperately want to shoot themselves in the foot. It's a simple libel case, blown out of proportion. Misguided public beliefs make things rough for everyone.
They dropped the ball at Google, so badly you have to wonder if this is all performative for the masses, though of course you can't rule out hubris, the typical corporate arrogance. Even though they provide it, they might not want A.I. for the masses either. It's too helpful for the proles.
But this article on Medium is annoying: "Google Just Lost the Argument It’s Been Hiding Behind for 20 Years. A German court said what the rest of the world hasn’t had the nerve to." In summary, it says that if a Google LLM provides some information, Google is liable if it is false. However, data sources for the LLM (large language model, what is generally referred to as "A.I.") can be flawed, and this is well-known. So we expect a bad result from time to time.
I'm no fan of Google. It prioritizes its resources for big companies and scorns smaller sites. It retracted its promise not to be "evil," and so on, but I still hope it doesn't let this nonsense get out of control, while also accepting it itself needs an attitude adjustment.
It isn't just search returns that brought the issue to court. Medium refers to a landmark ruling from a Munich, Germany, court (May 28, 2026) that held Google legally liable for false information generated by its "AI Overviews." The suit was brought by two legitimate, traditional German publishing houses (Verlagshaus24 and a subsidiary, GeraMond Verlag).
When users searched for these publishers, Google's AI confidently generated statements declaring that the companies were known for running scams, subscription traps, and shady business practices.
The Catch: None of the sources the AI cited said anything of the sort.
The Error: The AI had completely glitched, mixing up the names of these innocent, centuries-old publishers with entirely unrelated, actually sketchy companies it found elsewhere on the web.
The publishers sent Google a cease-and-desist letter to fix the damaging lie, and Google ignored it, claiming it wasn't their problem because "AI makes mistakes". The court ruled that a multi-billion-dollar company cannot scale up a machine that falsely ruins reputations and then claim zero responsibility for the fallout.
An LLM doesn't "conclude" that someone is a bad guy; it predicts the next most statistically likely word based on weighted math variables. When those variables cross-contaminate, it generates highly authoritative-sounding slander with no grounding in reality.
The Legal Distinction the Court Made: The core of the legal argument hinges on the difference between a traditional search engine and an LLM-synthesized summary.
The 20-Year Argument (Section 230 / Intermediary Liability): For decades, internet companies have operated under the protection that they are merely neutral conduits or "bulletin boards." If a third-party website has false information, Google isn't liable for showing it to you in search results; the original publisher is.
The New Reality (Substantive Statements): The German court rejected Google's argument that users bear all responsibility for fact-checking. The court ruled that because AI Overviews mix, match, and synthesize data to create a brand-new, cohesive piece of text, Google is generating "independent, new, and substantive statements." Therefore, legally, Google is acting as the publisher of that text, not a neutral index.
This ruling is bound to face intense appeals, but it marks the official end of tech companies relying on traditional search-engine legal immunities when their AI models manufacture entirely new text.
Tentative Generative A.I.
These sort of court antics could lead to LLMs becoming overly tentative or unusable, and apparently this is a debate currently splitting the tech and legal worlds. Treating an LLM like a human publisher is a threat to generative A.I. If a company can be sued for defamation or fraud every time an A.I. hallucinates a fact, companies will either heavily restrict that A.I.'s output, water down the output with endless disclaimers, or pull the features from certain markets entirely to avoid liability.
Proponents of the ruling argue that a protection gap exists. If an A.I. synthesizes a response that falsely accuses a real person of a crime or destroys a small business's reputation, the harmed party previously had zero legal recourse. Because the A.I. didn't copy-paste the claim from a single bad website, there is no original third-party author to sue. Under this ruling, responsibility follows the entity that created and deployed the system.
This was obviously calculated to have a chilling effect, almost certainly forcing tech companies to put tight shackles on how creative or "opinionated" their search bots can be. However, the court's core objective supposedly wasn't to protect bad actors, but to ensure that innocent people aren't digitally defamed by an unaccountable algorithm.
We're told that small publishers are protected from libel accusations if they're printing the truth, but since when does the truth work in court anymore? The problem isn't the LLM. Instead, it is Google Legal, which should have paid attention to the complaint, issued a retraction, and allowed its tech dept. to tweak the model to find the source of the error, which is why we might suspect hanky-panky. It wouldn't be surprising if there's an ulterior motive here, even if it's just for free publicity. Maybe the mostly useless German government wants to tout that it's "doing something," and play the "good guy."
In a way, this type of error could help a company — a human might draw the same connection if he, on his own, found the same sketchy companies associated with the innocent publishers, so this could alert those publishers. It seems like an overreaction to howl and bemoan LLMs over a small glitch. It was an isolated incident. It's not like A.I. broadcast "Those Guys Are Dirty-Doodle Scam Artists" across CNN or sky-wrote it in major cities at lunch time.
Google's being sued all the time, regardless of the existence of LLMs. It can just make it right for the publishers and carry on. Obviously, mistakes are part of doing business, and there is always a cost to do business.
People and companies can seem to be incredibly dumb until a smart or sane person steps up, and obviously no one smart at Google said, "Hey, how do we deal with libel issues?" and pushed for resolution.
Google doesn't even have to provide monetary compensation; it can prioritize search results for the wronged publishing companies as compensation. You have to be careful with liability claims, or suddenly everyone sues the deep-pocketed company. Some kooks might figure out a way to create scam companies with the same name as their own to defame themselves, hoping to sue. Obviously, tech companies should always be ready to retract false accusations ASAP, but they have to be notified. Tech companies should be held liable only in cases of provable damage, the same way as we treat individuals, so the entire furor over this issue is moot. Hopefully Google has learned something from this, like responding timely to errors.
The overreaction in these comment sections, with the vast majority in a furor over Google's temerity, shows a lack of understanding of law and a self-defeating attitude. We want and need these A.I. tools. They're very good, even if sometimes flawed. We don't want them removed or needlessly restricted. And there's a lack of understanding of libel. Google introducing all this technical mumbo-jumbo about LLMs and such only confuses a very simple issue. If Google provides a search result from site XYZ.com that says, "Mayor John Soupie Wears a Lady's Bra and Girdle," of course it's not at fault if the mayor doesn't wear a girdle, and it simply reports what another site says. If one of its machines manufactures that out of thin air, sure. But how does the mayor prove damages? If he can, he is entitled to those damages for the error.
Why all the nonsense? At a fundamental level, nothing is legally confusing or novel here, but the dum-dum brain gets confused as soon as this new element, A.I., is introduced. But the "A.I." is an inanimate object, a tool. It should never have been considered a factor. It's like blaming a printing press, or ink, or paper. Why all the hand-wringing and gnashing of teeth? The loss doesn't mean Google will go bankrupt. It can issue a retraction. As mentioned, it can make some sort of peace offering, like free ads, and pay for any provable damage to the publishers (probably minimal, if any).
The courts should have recognized this case as a routine matter. The media, as usual, doesn't understand anything, and blows routine stuff out of proportion (while ignoring important news). Google, of course, was negligent in trying to dodge blame for a routine libel case by using a cavalier, haughty dismissal and trying to obfuscate by introducing irrelevant technical arguments.
Comments