Meta uses our data to train its AI.

And basically all the headlines from the “cut and paste” traditional press are all the same: “Meta is about to use Europeans’ social posts to train its AI. Here’s how you can prevent it” (this one from Euronews).

It’s easy for our journalist friends on a fixed salary (on the payroll) to do their job: just repeat what others are saying and do not use your brain, or try to dig a little deeper into the facts.

Here, for example, is what an important radio station here in the french Riviera wrote this morning (28/05) on its site:

Meta, owner of Facebook, Instagram, and WhatsApp, can use posts, photos, and comments to train its artificial intelligence starting Tuesday, May 27th. This is legal under EU data regulations, but users can object by filling out a form via the privacy settings of each application. To disable this option, go to the privacy center on Instagram or Facebook (or use the form provided for WhatsApp), fill out the objection form, and you will receive a confirmation email.

They are wrong.

In our opinion, they are all wrong. Here is why.

Meta is the only big company to develop its AI in an “open source” way: datasets are known (as is evident here), the papers are public, and the models are open.

Local models

Yes: you can download Llama 4 (the name of the current model by meta.ai) and run it on your Mac (or a decently equipped PC) locally.

Without them—or anyone else—being able to know what you ask and what it (the AI) replies to you.

This is in stark contrast with, for example, ChatGPT / GPT-4.5 by OpenAI (and, as far as we understand, the amazing Gemini 2.5 by Google). We do not know where they got their training data from, and for sure, when we use their online AIs, they can trace each and every interaction we have.

Why not opt out?

So, for our part, we invite you… not to prevent Meta from training on your data.

In other words, let Meta train its meta.ai with your data.

Two reasons:

1. We would prefer that Meta advances rapidly with its open AI, rather than the closed suppliers. And training data is quite useful for what is called “pre-training” (sometimes post-training too).
2. If we wrote anything interesting, useful, or “correct” on Facebook or Instagram in the past… we think it is a good idea to let the AI learn from it.

Or, in other words, those who opt out are saying something like, “I do not want my opinions to be taken into consideration by future AIs, only those of other people.” Why anyone would wish that is totally unclear.

Meta has the data

And remember: Meta already knows all our data, or rather everything we’ve decided to write and upload to its systems (Facebook, Instagram, WhatsApp, Threads).

Your data, our data, your images, etc.: they already have it, and this opposition will change nothing.

If you really do not want third parties to know about you… you should not have used the social network in the first place.