S6, E247 - The EU's Bold Move on AI Training

Send us a text

The European Parliament has released a groundbreaking 175-page study concluding that AI companies' practice of training on copyrighted material without permission constitutes mass reproduction not covered by current laws. This study recommends transforming the landscape through an opt-in system, radical transparency requirements, and fair compensation models for creators whose work trains AI systems.

• EU study reveals AI companies are treating the internet like a free "all-you-can-eat buffet" of creative content
• Recommendation to shift from opt-out to opt-in system requiring AI companies to request permission
• Call for mandatory transparency about what data AI models are trained on
• Proposal for fair licensing models similar to Spotify where creators get paid when their work trains AI
• New EU AI Act regulations taking effect in August will incorporate some of these protections

Stay safe, stay informed, and always question the code.

Support the show

Episode Transcript

Speaker 1 (00:00):
Is your favorite AI creating art from stolen goods?
It's a huge question nobody wasready for, but Europe is
tackling it head on.
Welcome to Privacy, please,where we decode the digital
world for you.
I'm your host, cameron Ivey,and today we're diving into a
bombshell 175-page study fromthe European Parliament that

(00:22):
could change the future of AIforever.
So you know those amazing AItools that can write a poem,
design a logo or even generate aphotorealistic image of your
dog on a moon.

(00:42):
They seem like magic, buthere's the not-so-magical secret
.
To learn how to do that, theyhave to study A lot.
And what are they studying?
The entire internet thatincludes your favorite artist's
profile portfolio, thephotographer you follow on
Instagram, best-selling novelsand investigative journalism.

(01:05):
They are, in essence, readingeverything.
The big problem they're notasking for permission.
This new EU study basically sayshold on a second.
They've concluded that the waythese AI models are trained on
copyrighted material is a formof mass reproduction, and the

(01:28):
current laws just aren't builtfor this.
Just imagine a robot grabbingbooks and art and photos from a
digital library.
So what did the EU's deep diveuncover?
175 pages is a lot.
If you don't have the time, I'mgoing to break it down for you
into three main bombshells.
Number one the all-you-can-eatbuffet is over.

(01:51):
Right now, ai companies areessentially treating the
internet like a freeall-you-can-eat buffet.
The EU has a rule called thetext and data mining, or TDM,
which was meant for research,but this study shows training a
massive commercial AI model isnot the same as academic

(02:12):
research.
The big recommendation flippingthe script from an opt out
system to an opt in.
Think of it like this Opt outis like a restaurant
automatically adding a 20% tipto your bill and you have to
notice it and ask to have itremoved.
Opt-in is when they ask youfirst if you'd like to leave a
tip.
The EU is suggesting that AIcompanies should have to ask for

(02:36):
permission to use creative workfor training, not just take it
until someone tells them to stop.
This is a game changer forcreators.
Number two we need to see thereceipts Right now.
What these AI models aretrained on is a total black box.
It's a secret recipe, but whatif that recipe includes your

(02:59):
private photos, biasedinformation or pirated content?
The study demands radicaltransparency.
It proposes that AI companiesmust provide a detailed summary
of what their models have beentrained on.
This isn't just about copyright.
It's a massive win for privacyand fighting bias.
It means we could finally seeif our personal data is being

(03:22):
used to train the next big AI.
And finally, number three, fairpay for fair play.
If an AI learns its style fromthousands of artists, shouldn't
those artists get a piece of thepie?
The study says a resounding yes.
It calls for a new and fairlicensing models.
Imagine a system, maybe likeSpotify for AI training, where

(03:46):
creators get paid every timetheir work is used to teach a
machine.
This could ensure that humanswho create the foundational
culture and knowledge of AI areactually compensated.
Weird.
So what does this all mean foryou on your phone, right now, on
your computer?
This is more than just a legaldocument.
It's a battle for the future ofcreativity and, yes, your

(04:10):
privacy.
The EU's AI Act is alreadyputting some of these ideas into
motion, with new rules set totake effect in August.
This study will fuel the firefor even stronger protections.
It pushes for a world where AIinnovation doesn't come at a
cost for human creators.
It's a call for an AI that isnot only smart but also fair and

(04:31):
transparent, and in a worldwhere our digital lives are
constantly being mined for data.
That's a privacy conversationwe all need to be a part of, so
next time you ask an AI tocreate something for you, think
about what it learned fromThanks to the EU.
The answer to that questionmight soon be public knowledge.
That's all the time we have forthis episode of Privacy.

(04:54):
Please.
I'm Cameron Ivey.
Stay safe, stay informed andalways question the code.