Meta Begins Training AI on EU User Data Through GDPR Loophole

Meta Begins Training AI on EU User Data Through GDPR Loophole

Redacto
6 min read

Meta (the parent company of Facebook and Instagram) has begun roll out of its AI features in the EU.

Meta says this will improve tools like Meta AI and Imagine, as well as future integrations across WhatsApp, Messenger, and Threads.

Watchdogs are on the defensive, with NOYB sending a cease and desist and widespread criticism in the EU. NOYB’s concerns largely focus on Meta’s decision to default users to opted-in, allowing AI training on their data. Instead, Meta should be asking users for their explicit consent before any opt-in or data collection.

Privacy organizations are well-grounded in their concerns given Meta’s history of knowingly training AI on pirated literature.

What Data Is Meta Collecting for AI Training?

Meta has confirmed that it will train its AI on publicly shared Facebook and Instagram content – excluding private messages, non-public posts and content from minors. While the concessions made to Meta’s data collection are a step in the right direction, they’re no substitute for informed consent from the wider base of EU Facebook and Instagram users.

Data included:

  • Text from public posts and comments
  • Image captions and hashtags
  • Your name, public bio, and public profile details

This means your cat meme from 2017 could be helping Meta generate content or refine its chatbot responses today. A public comment you left 20 years ago that might no longer reflect who you are could be crystallized in the training of future Meta AI models.

⚖️GDPR vs Meta: A Legal Standoff

Meta’s legal basis is that they claim “legitimate interest” in using the data to train their models under the GDPR.

But critics – notably NOYB (None Of Your Business) – argue this violates EU law, because:

  • Legitimate interest must be “necessary,” and;
  • Less invasive alternatives (like asking for consent) exist

NOYB have filed 11 formal complaints across the EU, warning that Meta’s approach disregards basic GDPR principles by assuming user participation.

Meanwhile, a German court sided with Meta, stating that scraping public data isn’t automatically a GDPR violation – so long as opt-outs and transparency measures are provided.

While this requirement is met, there is no substitute for obtaining informed consent from individual users. Meta allows opt-outs within a few clicks; the infrastructure exists to take a consent-based approach – but Meta know this will result in a smaller corpus of training data.

❌Can You Opt Out of Meta AI?

Yes – but the system is designed to encourage you to skip past the opt-out so Meta can maximize the data they collect.

From today, users in the EU should start getting prompted about Meta’s AI features and training. This prompt should be an opt-in request, but it’s pitched as informative (see: propaganda) by Meta, with a “link to the opt-out form”.

We’d recommend opting out as soon as possible – once you get the prompt, Meta are primed to start harvesting and training with your data.

How to Opt-Out of Meta’s AI training on Facebook:

  1. Log into your account
  2. Click your profile, then “Settings and privacy”
  3. Click the “Privacy centre”
  4. Expand the section “How Meta uses information for generative AI models and features”
  5. Scroll and click on “Right to object”
  6. Submit your email address

How to Opt-Out of Meta’s AI training on Instagram:

  1. Navigate to your own Instagram profile while logged in
  2. Tap the expandable three-line menu and choose “Settings and privacy”
  3. Find “More info and support” then “About”
  4. Tap “Privacy policy”
  5. Tap “Learn more about your right to object”
  6. Submit your email address

If you don’t do this, you will be opted-in by default.

🎯Meta’s Framing vs Reality

Meta presents this initiative as “AI for Europe” – localized, culturally aware, and innovation-focused. Their PR says it’s about “helping people with creativity and productivity.”

Propaganda is defined as information, ideas, opinions or images, often only giving part of an argument, that are broadcast, published or otherwise distributed with the intention of influencing opinions.

Meta’s decision to present AI features through ‘informative prompts’ along with selectively publishing about the economic benefits of AI is difficult to describe without using the word propaganda.

Here’s the reality of Meta’s AI training approach;

This isn’t AI for Europe. It’s AI from Europe – built on your digital footprint, harvested without explicit, informed consent.

🔐Why This Matters for Digital Privacy

Meta has continued to set dangerous precedents for the development of AI. If they can get away with using GDPR loopholes to train LLMs, there’s very little stopping other technology giants from doing the same – and generally, they jump at these opportunities;

As generative AI expands, public data becomes a goldmine for machine learning – and a minefield for user privacy.

🛡️Final Thoughts: Know Your Rights, Guard Your Data

Meta’s rollout of AI in the EU is a bold (and arguably reckless) attempt to squeeze value from users without informed consent, aided by propagandist tactics.

As the saying goes – “If you’re not paying for the product, you are the product.”

Now, not only your attention will be sold off to the highest bidder, but your content and data is likely be handled the same way.

Our recommendation – opt-out. Review your privacy settings, and stay informed. With the rapid growth of AI, unethical training, and hunger to pad their bottom lines’, big tech are not slowing down.

Meta is just one company building AI – if you’re looking for help opting out of AI features and training for all major platforms, check out our AI Opt Out guides here.

You might be skeptical of simply opting-out, and prefer to simply withdraw as much content as possible from Meta platforms (or others). Redact.dev is the easiest, most comprehensive mass deletion tool available – you can bulk delete your content from almost all major platforms, and automate ongoing deletion to keep your digital footprint and AI training contribution minimized.

© 2025 Redact - All rights reserved