Someone Googled “Jimmy Keene net worth,” hoping to figure out how much money the author, TV producer, and former FBI informant has tucked away after his memoir became Apple TV’s Black Bird. Google’s AI Overview explained that Keene was “serving a life sentence without parole for the murders of three women.”
Google doesn’t think it should have to face defamation claims over this summary. The federal courts disagree.
It’s easy to see where the bot screwed up. Jimmy Keene went to prison on a marijuana charge, but took an FBI deal to transfer prisons and befriend a suspected serial killer Larry Hall. Hall eventually confessed to Keene and showed him a map of where he hid the bodies. Hall’s appeal failed and the government commuted Keene’s existing sentence.
From ‘Associate’ To ‘Doppelganger’: The Evolution Of The Legal AI Persona
Confronting Biglaw’s ‘snake eating its own tail’ problem.
So Google’s AI gathered a bunch of search results about Keene gathering evidence on a serial killer and reported it out as though Keene was a serial killer. It’s not clear where the bot came up with “three,” because that matches neither “dozens,” the number of murders Hall committed nor “zero” the number of murders Keene took part in.
Longtime readers know my fondness for the Christine Lemmer-Webber description of AI as “Mansplaining as a Service” — it’s often wrong, but always confidently wrong — and now it’s come for Google’s AI summaries.
Keene sued for defamation, and last Monday, Judge Thomas Durkin of the Northern District of Illinois denied Google’s motion to dismiss as to the two overviews that fingered Keene for crimes he didn’t commit.
Google, represented by Williams & Connolly, tried to sink the case by arguing that no one really believes Google’s AI summaries anyway. Those summaries are the new fresh hell that demotes actual Google search results below a bot-generated SparkNotes version. Correction: the summaries demote the several spammy “sponsored” results, which had already themselves demoted the actual search results. In any event, the Google summaries are so bad that the other day I had to go to Bing to verify a basic fact. BING!
The Lateral Market Runs On Recruiters. goLateral Wants To Cut Them Out.
goLateral is a quiet window into what's actually out there. Joining takes two minutes, costs nothing, pays you $100, and stays completely confidential.
We used to be a proper society.
To be clear, people shouldn’t trust those overviews. We already covered Google’s AI making up fake state ethics rules, and nothing has happened in the interim to suggest it’s solved its problems. The problem with this defense is that Google represents to its users that they should be able to trust its output and it’s hard to rug pull that at the last second.
Google’s defense asserted in its brief in this case that Google search exists to “connect a user most efficiently to relevant and reliable information.” Which is true! But Judge Durkin pointed out that the company wasn’t giving the user that “relevant and reliable information,” but was instead downgrading that “relevant and reliable information” behind AI slop.
From the perspective of an ordinary user, asking an AI LLM to generate work product and receiving numerous indicia of falsity is different from a basic Google search. In Google’s own words, a Google search is meant to “connect a user most efficiently to relevant and reliable information.” R. 14 at p. 3. Although Google has now developed its own built-in AI LLM to “enhance” a search by summarizing information, id., an ordinary user searching “Jimmy Keene net worth” on Google is seeking factual information, not AI work product.
To get around this, Google claimed that its summaries include a disclaimer that they “may include mistakes,” but the court noted the disclaimer appears nowhere in the complaint or any exhibits. In fact, Keene’s lawyers say their own test searches never turned up this disclaimer either. This may be a factual fight for later, but it’s in enough dispute to survive the motion to dismiss.
Apparently, each of the offending summaries cited Keene’s Wikipedia page, which does not — at all — claim that he murdered three people. Including a verifying link can serve as a defense to defamation where commentary or opinion may be misleading and the link would clarify the underlying facts for the audience. But Judge Durkin distinguished the cases Google cited for this premise.
By pointing out that the statements in the AI Overviews are objectively verifiable, Google concedes the very point it is arguing against — the statements are facts.
It’s a weird flex for a product explicitly marketed to prevent the user from clicking. The problem is that If Google wanted users to click Wikipedia, it would just put Wikipedia as the top result. The summary exists to signal to the person searching that they don’t need to click on anything. Google did argue that. Users could have verified the truth by clicking through, it said. Set aside the audacity and there’s an economic problem. AI Overviews exist so you don’t have to click. That’s the product — the answer sits at the top of the page and the trip to the source becomes optional, as every publisher watching its traffic can confirm. Google’s litigation position depends on users doing the one thing its product is designed to make unnecessary.
And, not for nothing, but making the uniquely heinous claim that someone is a multiple murderer introduces heightened defamation issues that the “just have a link” cases never addressed.
Keene is a limited public figure, so he would need to clear the actual malice hurdle. But he claims he complained to Google three times — the same number of times he did NOT murder anybody — including through Google’s “report a legal removal issue” form. The court pointed out that these emails leave Google in a bind:
Either this report made its way to an individual at Google, which could plausibly support actual knowledge of falsity, or no human at Google reviewed the legal removal request and subsequent correspondence, which could plausibly support reckless disregard for falsity.
Google built the gold standard in putting answers in front of eyeballs. Now it’s hiding them under a summaries that they promise usually don’t mistakenly brand someone a triple murderer. That’s not good enough at the motion to dismiss stage.
Joe Patrice is a senior editor at Above the Law and co-host of Thinking Like A Lawyer. Feel free to email any tips, questions, or comments. Follow him on Twitter or Bluesky if you’re interested in law, politics, and a healthy dose of college sports news.