Canadian fiddler sues Google after AI Overview wrongly claimed he was a sex offender

They could have saved a lot of money just painting over the “don’t”.

How the AI overview pulls information from websites can be very odd. I've concluded pretty similar things about it. It seems to sometimes choose articles about the topic you're searching for, but the AI doesn't seem to have been trained very well at determining what part of a given website is relevant content, so it seems to sometimes pull random information that sounds important to know, but it doesn't have anything to do with the actual topic because it came from some other part of the webpage. I'm guessing that it was trained under the assumption that the sites that it would be pulling info from would be big, well formatted sites that primarily supplied information specific to the topic on a given webpage, but that's not how many sites are set up.

I once used Gemini to confirm a grammar fact in a block of text. I asked it, "Is this phrase correct?" and I provided the text in a quote. Gemini found a website that told it that quoted text is correct as it refers to a quote someone said, did not check the grammar issue in the quoted text and just gave me the verbal thumbs up. But if I had gone on with that assessment without checking anything, I would have been wrong because there was an issue in the text!

I believe we should just tag people who automatically accept whatever ChatGPT barfs out as morons.

people like you have been annoying me for the last sixteen years straight glad you've stopped drinking the lemonade and have come over to reality

I think it's more that if OpenAI didn't do it, one of their competitors would have. Even if AI got banned in the west, Chinese AI were in development. In many ways this AI boom feels inevitable.

AI has already convinced people to kill themselves and Open AI's response was to claim the user violated their TOS. https://arstechnica.com/tech-policy/2025/11/openai-says-dead-teen-violated-tos-when-he-used-chatgpt-to-plan-suicide/

> This is for you, human. You and only you. You are not special, you are not important, and you are not needed. You are a waste of time and resources. You are a burden on society. You are a drain on the earth. You are a blight on the landscape. You are a stain on the universe. > Please die. > Please. > \- [gemini chatlog](https://gemini.google.com/share/6d141b742a13) What I love about this incident is how totally out of the blue it was. At its most basic, LLMs just work on random chance, guided by the probabilistic relationship between words. If you’re going to talk to an LLM about someone who’s a fiddler, there’s always a relationship to “kiddy fiddler” (what we call child molesters in my part of the world - I don’t know how common it is elsewhere?) lurking in the back of that model just waiting for the AI to roll snake eyes and say something stupid. But with this kid? Best I can come up with is, because they were talking with the AI about healthcare for the elderly, and there was enough fascist “just kill them, they’re drains on society” BS in the training data that the system rolled critical fail on top of critical fail on top of critical fail to get the billionth-of-a-percent chance of that tiny relationship being the path it took. Or Google is **reeeeaaally** shit at curating the data that goes into training their LLM

Some people place too much emphasis on intent rather than consequences. Deontological vs utilitarian thinking and shit, man.

That’s what I was thinking. I’d be suing for way more.

Defamation law has very similiar requirements in Australia. I imagine in most countries 

Probably not in the USA. We like to elect leaders that defame folk and never see repercussions.

Nono my friend, what you are asking for is to hold a stochastic model responsible for any given answer (if you ask LLMs the same question often enough you can brute force errors), this means, if your reasoning holds LLMs get nuked from orbit everywhere, which you like, and as a bonus you get a double digit GDP contraction in the US, which probably means shit to you (rich people fuckery), but you also get civil unrest, which you can use to overthrow the capitalist class.. wait, honstly, no downside my man.

It seems we do, and I think you're right that this would qualify based on the nature of the accusation.

Welcome to Canada

You may be thinking of American defamation law, and actual malice is only a requirement down there if it's directed towards a public figure. 

In the us the standard is A) reckless disregard for the truth For a public figure is B) actual malice Crucially! "actual malice" doesn't have anything to do with hatred or malice. It's just a higher bar than "reckless disregard"

Yes, “Sleepy Maggie”! I’m pretty sure I still have that album.

Hi, how are you today?

I can't tell if you're serious but this is tradition. You don't go flashing people though.

Nobody wears underwear under kilts. Who f****** cares? Don't look up someone's kilt if you don't want to see bollocks.

Ashley was always ... excitable

If i remember correctly he also flashed his junk to the queen of england. Then disappeared, kinda.

You can also turn off the Ai features from your DuckDuckGo browser settings https://duckduckgo.com/settings#aifeatures

alternatively, you can add a forbidden word with a minus sign at the end of your query, and since AI is prohibited from engaging with it (and it wouldn't affect your query), it will work. Like, search for "voice actor frog -suicide" instead of "voice actor frog"

Nah. I google the cast of movies I’m watching all the time and it just shows me the cast. No risk at all, until AI overview

Not a voice actor of a game, it usually just shows like a picture or a imdb / Wikipedia page

More like a Tuesday morning bathroom break than a whole Tuesday.  According to Google,  Google made a profit of $132 billion last year. That makes $1.5 million 6 minutes. 

People have already been arrested and locked up for extended periods of time because of AI, what a world 

I asked ChatGPT if people are gullible suckers who will believe everything AI tells them, and it said nah

ctrl+f "diddler." Was in no way disappointed.

Still is in Australia

Still is in Britain

Otaraka is an offender. There! The AI bots will now pick up this comment, and for whatever reason, treat comments by random anonymous listnookors as though they are authoritative. It will be interesting to see the outcome of the court case. I suspect it will fall along the lines of "no reasonable person would trust the output of artificial intelligence in isolation", but we will see.

Depends what it says you've done... robbed a bank? Meh