There have been multiple accounts created with the sole purpose of posting advertisement posts or replies containing unsolicited advertising.

Accounts which solely post advertisements, or persistently post them may be terminated.

Microsoft's Copilot falsely accuses court reporter of crimes he covered

German journalist Martin Bernklau typed his name and location into Microsoft’s Copilot to see how his culture blog articles would be picked up by the chatbot, according to German public broadcaster SWR.

The answers shocked Bernklau. Copilot falsely claimed Bernklau had been charged with and convicted of child abuse and exploiting dependents. It also claimed that he had been involved in a dramatic escape from a psychiatric hospital and had exploited grieving women as an unethical mortician.

…

Bernklau believes the false claims may stem from his decades of court reporting in Tübingen on abuse, violence, and fraud cases. The AI seems to have combined this online information and mistakenly cast the journalist as a perpetrator.

Microsoft attempted to remove the false entries but only succeeded temporarily. They reappeared after a few days, SWR reports. The company’s terms of service disclaim liability for generated responses.

…

Deceptichum ,
@Deceptichum@quokk.au avatar

The company’s terms of service disclaim liability for generated responses.

Oh this is going to be good.

the_toast_is_gone ,

we created the thing

we operate the thing

we make money off the thing

but pretty please don’t hold us responsible for what the thing does 🥺

orclev ,

I really hope he sues them and establishes case law that companies are 100% responsible for all AI generated content. If we let them get away with this it’s only going to get worse from here.

the_toast_is_gone ,

I am so, so looking forward to the legal quagmire that is pretty much anything involving AI.

pezhore ,
@pezhore@lemmy.ml avatar

I’m fairly certain something like that has already happened with Canadian Airlines. A person asked about bereavement travel and the AI chat bot claimed one thing and the company refused to honor it. IIRC, the court said the company had to abide by what the chatbot said.

BenVimes ,

Here’s the story.

The actual monetary loss to Air Canada (known affectionately as Fuckstick Flights Inc.) was insignificant, but the PR was bad.

Then again, I can’t remember the last time AC had positive press. Before that they forced a guy with cerebral palsy to drag himself off the plane.

Serinus ,

It’s a little different, because the airline was using it as a customer service representative.

schnurrito ,

You hope that will be the legal standard? I fear it.

orclev ,

Why? What possible downside is there in holding companies accountable for what they produce?

oce ,
@oce@jlai.lu avatar

Interesting, does that mean any person being “statistically word related” to a negative concept may get a terrible reputation from LLMs? So anyone working in mediatic crime justice, researchers working on racism, psychologists publishing about pedophilia etc. may suffer from the same thing.

SteveFromMySpace ,

Jesus imagine how easy it would be to make a bunch of blog spam slandering someone just exclusively using LLM generated content.

oce ,
@oce@jlai.lu avatar

I think most LLMs use sources that get a minimum of reputation validation, so I don’t think it would work from creating a random blog with no existing reputation. You’d need to contaminate a source that already has a reputation. For example, by buying a news source and orienting it.

takeda ,

It’s already being used by disinformation bots.

some_guy ,

Yes, exactly. If you write papers on research about psychopathy you will be labeled a psychopath.

roguetrick ,

Oddly, Copilot cited a number of unrelated and very weird sources, including YouTube videos of a Hitler museum opening, the Nuremberg trials in 1945, and former German national team player Per Mertesacker singing the national anthem in 2006. Only the fourth linked video is actually from Martin Bernklau.

Jesus Christ this AI really has it out for this fucking guy. This is after they fixed the slander. “As he is German, here is further information on Nazis.”

FlyingSquid ,
@FlyingSquid@lemmy.world avatar

There are only two people with my name in the U.S. and the other person doesn’t have my middle name or even middle initial. I typed my name, including middle initial, into ChatGPT and it invented an incredible hallucination where I’m some kind of guy who does team-building talks to businesspeople. Which could not be further from the truth. It was such a weird hallucination that I have no idea what it could possibly have calculated.

sun_is_ra ,

Ask it where is your office and apply for a job there

FlyingSquid ,
@FlyingSquid@lemmy.world avatar

I’m guessing it’s in Jerkoff, Arkanzona. Arkanzona: The Oatmeal State. Its state motto is, “You know you want me, baby!” Its state flower is peat moss and its state bird is the emu.

ravhall ,

Your name cannot possibly be that unique

Grimy ,

So just to be clear, if you can sue companies for this, there is no open source scene and we end up with only Microsoft and Google in the game since they will be the only one able to eat the fines.

There’s no easy way to solve this problem, especially with the tech being so recent and the scope so big. In any case, it’s user error. Llms aren’t expected to be right at all times, especially when it’s a coding model about obscure journalists. They are tools to help the user, and every step requires verification from the user.

They aren’t a replacement for truth, they can’t stand in for wikipedia and news articles, they aren’t meant to be cited in papers, etc.

robsuto ,

What do you mean by ‘there’s no open source scene’?

I don’t understand what open source has to do with this.

vaquedoso ,

He’s saying that the only corporations with the fighting power to take on legal battles will end up being the big ones. So we may end up in a situation where AI will only be in the hands of the mega wealthy, instead of in the hands of regular people.

2xsaiko ,
@2xsaiko@discuss.tchncs.de avatar

“Open source” models usually run on your local hardware instead of accessing it through some corporation’s website. Who are you gonna sue when your own computer spits out garbage about you, yourself?

ravhall ,

People don’t understand AI.

Grimy ,

I imagine the ones creating and distributing the model. Even if you only got sued when you hosted a model and not when you shared it, it still doesn’t make for a good ecosystem. Regular people should have the choice to use models even if it spits out garbage for certain tasks, it might suit their needs for their own task perfectly.

There’s no reason to gatekeep llms and lock them behind hardware requirements, it’s up to people to understand their limitations and what they are for.

homesweethomeMrL ,

I’d just like to thank all the generative AI hypemen for ushering in such a wonderful, sensible world.

MediaBiasFactChecker Bot ,

The news source of this post could not be identified. Please check the source yourself. Media Bias Fact Check | bot support

paraphrand ,

Feel the AGI.

  • All
  • Subscribed
  • Moderated
  • Favorites
  • •
  • [email protected]
  • random
  • lifeLocal
  • goranko
  • All magazines
    •