Meta won't train AI on Euro posts after all, as watchdogs put their paws down

Meta has caved to European regulators, and agreed to pause its plans to train AI models on EU users' Facebook and Instagram users' posts — a move that the social media giant said will delay its plans to launch Meta AI in the economic zone. For everyone else outside the EU, Meta will be going full steam ahead using your public …

[heyrick]

"further delays bringing the benefits of AI to people in Europe."

Please, ban even more. We don't need AI for AI's sake.

Maybe, by the time it rolls out in Europe, few will bother as it will by then have been exposed as the giant chattering-parrot pile of bovine excrement that it is.

[ecofeco]

Re: "further delays bringing the benefits of AI to people in Europe."

This.

The average person is not going to benefit from AI in any way.

It WILL be used against them. That is what rich people DO.

[SundogUK]

Re: "further delays bringing the benefits of AI to people in Europe."

You don't actually know any rich people, do you?

[Anonymous Coward]

Re: "further delays bringing the benefits of AI to people in Europe."

well my heads not up their arse like some (look in a mirror when you remove it)

[Snake]

Re: rich people

I do. And while they are not the sociopaths that we'd all to identify them as, what they *do* do is see [their] entire world as transactional: benefit versus costs. Even their personal relationships.

The moment that they no longer believe that a transaction or even a personal connection is strictly to their benefit - BAM, it's over. You'll get dumped like last week's rotten meat. If you'd like to believe otherwise then you don't know many rich people.

[Michael Wojcik]

Re: rich people

Well, that is rather a generalization, and it does depend very much on what you mean by "rich".

My household income is in the top 1% in the state where I currently live; it was in the top 1% in the state where I previously lived. It is very, very far from the median for the top 0.1%. I rarely worry about money. I have ready access to a basket of resources that for nearly all of human history would have been ridiculously generous. Am I rich?

I have friends who have more wealth (though not, in every case, more disposable liquidity) than I do. Are they rich?

I don't "see [my] entire world as transactional". Nor, I believe, do my friends.

Now, do I trust anyone in a position of power at Meta, or at OpenAI, or at Google, etc, to have my interests at heart, or the interests of most other people? No, I very much do not. But I'm reluctant to claim that correlates with being "rich" in some sense, much less see the latter as causal.

[Anonymous Coward]

Re: rich people

transactional?

in my opinion that makes them sociopaths

[Richard 12]

It's illegal, Meta

It doesn't matter what your business case is. It's illegal.

If my business relied on my employees wandering into Zuckerberg's house and stealing his stuff, you'd quite rightly call the police.

Why are you surprised when we do the same?

[OhForF']

Re: It's illegal, Meta

Meta is paying a lot for his security so police would only be called to collect the remains of your "employees" after they were gunned down by the "security force". Mark would probably be happy to go back to wild west times with less law enforcement as he would be able to get away with way more shenanigans than today.

[Bartholomew]

Backroom thinking at facebook

You can bet your last dollar that Facebook are now thinking of something like this:

OK we legally can not train any AI on the data and information directly provided by humans in Europe. But ... now hear me out ... what if we train the AI on the metadata about our user in Europe, and we anonymise the source of the metadata could we sidestep any toxic GDPR fallout that way ? This way we may still manage to gain aggregated actionable information for advertisers that is at worst with the granularity of a city or region. But within Europe that may be more than our competitors, and that just may be enough to bump the bottom line.

[0laf]

Re: Backroom thinking at facebook

I'd suspect it's more like, "We've already used the data to train it but we need more. How do we not get caught?".

[Filippo]

And once again Meta proves utterly unreliable on critical privacy issues. It's obvious that they don't give a crap about the law; their only concern is whether they'll get caught.

BTW, I used WhatsApp back then, and I requested account deletion, and sent a GDPR request to delete all of my data, when Meta announced they would transfer WhatsApp data to Facebook. Never been honored. People are still able to find me on WhatsApp. They just break the law all the time, and get away with it, and that's all that matters to them.

[Anonymous Coward]

Claims versus what they do

Even if they promise something, you often don't get what you'd think they promised.

Take the original way they were going to use all European user's (and everyone visiting one of the copious sites containing their illegal tracking scripts) data for their AI training. They used the "legitimate interests" clause, that is abused to try and get away with anything that is illegal under GDPR. (thanks to industry lobbyists \s). There you can not deny but only object.

So reread Meta / Facebook's original option. They'll use your pictures, conversations and all other data they have on you. If you do not agree, you can object. The process is obscure and requires you to write "why and how you feel this legitimate use by Meta harms you". Read again: you can *object* NOT *deny*. In other words, 99% chance that the original plan was to continue to use your data for AI training even if you object since objecting is legally not the same to denying and they denied you the option to deny.

[Dan 55]

Re: Claims versus what they do

"why and how you feel this legitimate use by Meta harms you"

This description from Meta is bollocks. You don't need to go into the whys and wherefores of how you feel, all you need to do is say the data does not fall under the definition of legitimate use under 6(1)(f) of the (UK) GPDR and you're invoking your right not to consent for your data to be processed in this way under article 4(11).

[Filippo]

Re: Claims versus what they do

You can, but I did something similar when I wanted to get out of WhatsApp, and they outright ignored it. Chances are they'll ignore individual complaints. They know regulators don't go after individual complaints anyway, and most of us won't actually hire a lawyer to fight them.

Even if they don't ignore it, and promise you they're doing what you asked, there's no way to verify. And, because they've already proven over and over again that they will break the law and/or lie if they think they won't be caught, they can't be trusted on that.

[Anonymous Coward]

Re: Claims versus what they do

They're welcome to train their AI on my Whatsapp messages, which I can assure the world will result in no worthwhile "learning" of any kind. It will however feed the sort of outputs that already get AI a very bad name.

[Mark #255]

Re: Claims versus what they do

I wonder what they'll do with my daily Wordle* score...

* Other daily word quizzes are available

[Christoph]

Re: Claims versus what they do

You don't even need that much detail. The reply accepting my objection was so quick that they may have a bot looking for the string "data protection"

[Snake]

I'm going to get flamed for this

"And once again Meta proves utterly unreliable on critical privacy issues."

What privacy? According to the story and Meta's claims, they will be using your PUBLIC social media posts for AI training.

Public. You intentionally made a post of your thoughts that you allowed anyone and everyone to see, by selecting "public", but now you want to select what 'public' really means.

Want privacy?? Stop posting publicly on social media!!

[Anonymous Coward]

UK?

The statements quoted imply this affects the UK too, but the article talks about the EU.

Does it cover the UK or not?

Whilst it will be great if it does, expect the usual Reg nutters to moan about not being the fictional brexit they voted for!

[Anonymous Coward]

Re: UK?

LMGTFY

[Anonymous Coward]

Re: UK?

Yes, whilst I could have googled that myself, my comment was more alluding to the fact that this should have been made clear in this article itself.

Thanks, though, you did save me some time doing that :-)

[anothercynic]

Re: UK?

I wouldn't bet on it covering the UK. We are after all no longer in the EU.

[Anonymous Coward]

Re: UK?

It turns out, it does cover the UK.

Thanks to the anon coward above for the link: https://lifehacker.com/tech/opt-out-of-meta-instagram-and-facebook-ai-data-training

[JohnMurray]

Does it cover the UK or not?

Probably not.

Even if it does, meta will ignore it until they are dragged sc reaming to court, and even then they'll ignore it [out govt/s are pussies]

[Coastal cutie]

Re: Does it cover the UK or not?

It seems to - I'm a UK resident and was able to object

[0laf]

Re: Does it cover the UK or not?

I objected to but got the feeling Meta was crossing its fingers behind its back when it said "pinky promise we won't slurp your data"

[Anonymous Coward]

Re: Does it cover the UK or not?

It turns out, it does cover the UK.

Thanks to the anon coward above for the link: https://lifehacker.com/tech/opt-out-of-meta-instagram-and-facebook-ai-data-training

[Tron]

AI will be Americentric.

Although training an AI on social media posts is a really bad idea, they have to train them on something. If they lock out European data, AI will be Americentric to its core.

The EU is progressively locking out an awful lot, from low cost Chinese EVs and websites to consumer products and food. The EU may not be as self-sufficient as it thinks it is, and isolation merely breeds xenophobia..

[Dan 55]

Re: AI will be Americentric.

AI will be Americentric like all the rest of the Big Tech products over the past 40-odd years (no localisation, bad localisation, US results favoured over local results, etc...). So what? The threat of "give us your data or get Americentrism" is an empty threat, the rest of the world is already used to it.

[Watashi]

Re: AI will be Americentric.

True that. When I was a Uni student in the 90s we called this "American Cultural Imperialism" and treated it like a bad thing that should be stopped... but these days the Americancentric world-view has become so normalised accross the whole of UK society that even the most bolshie students of today have become thoroughly Americansied.

[Anonymous Coward]

Re: AI will be Americentric.

so it's AI will imitate an an average american, so technically it won't be an AI.

It will be AS, American Stupidity in concentrated form.

[Nematode]

...agreed to PAUSE its plans...

Nuff said

[Nematode]

The biggest elephant in the room is the value of any AI trained on the cesspool that is Mera's products. Bias towards total idiocy and influencers.

[Anonymous Coward]

> on the cesspool that is Mera's products

Careful now, Aquaman does have quite a temper and you don't want to be having a go at his wife.

[ecofeco]

And the EU can verify this...

...how?

Seriously, does anyone believe this?

[spireite]

Garbage in. garbage out.....

Having experienced LLM being used on data in my work sphere, never before has the old adage 'Garbage in, Garbage out' been more fitting.

It goes without saying that Facebook is 99% garbage......

[EvaQ]

"Meta won't train AI on Euro posts"

... so how about Pound posts?

[LybsterRoy]

The good news: reducing the amount of Facebook posts used to train an AI MAY result in increased sanity of said AI

The bad news: people think their facebook posts are worth protecting because "private"

[Potemkine!]

OMG, how will we be able to live without that fantastic experience? Without it, civilization as we know it will probably collapse, people will begin to eat each others and stars will fall from the sky.

[Northern Harrier]

Yet another

Brexit benefit.