Meta AI

Objecting to Meta AI training: where the official route is

Meta has said it trains its AI models in the EU on public content shared by adults on Facebook and Instagram. Meta offers a form on each platform through which people can object. This page shows where those forms are, what Meta and the consumer advice centres say about them, and what stands where, each with its source and date. Status: 18 October 2026. It does not say whether anyone should object.

What Meta says it uses

In its announcement of 14 April 2025, Meta writes that it trains its AI models on public content, like public posts and comments, shared by adults on our products in the EU, and on the interactions people have with AI at Meta. Two limits are stated in the same article: We do not use the content of your private messages with friends and family to train our AIs, unless someone in the chat chooses to share those messages with Meta's AI features (a clarification added on 27 March 2026), and public data from the accounts of people in the EU under the age of 18 is not being used for training purposes.

Sources: Meta, “Making AI Work Harder for Europeans”, 14 April 2025, updated 27 March 2026; Meta Privacy Center, generative AI. Read on 18 October 2026.

The official forms

• Instagram: objection form on Instagram Help.
• Facebook: objection form on Facebook Help.

According to the Verbraucherzentrale, both forms are only shown to people who are logged in and are available only to people in an EU country. An e-mail address is required so that Meta can confirm the objection; a reason is not required. If the links do not open the form, the same page describes the route through the platform's privacy policy. On 18 October 2026 both addresses answered with HTTP 200 and asked for a login.

What an objection covers, according to the consumer advice centre

The Verbraucherzentrale (page dated 23 June 2025) writes that Meta has used public content in the EU for AI training since 27 May 2025. Whoever had not objected by then must assume their data is used. An objection remains possible afterwards, but kann jedoch nur für die Zukunft berücksichtigt werden (it can only be taken into account for the future), and posts published after an objection dürfen nicht mehr in die KI einfließen (may no longer flow into the AI). Data that has already been used cannot be retrieved. The objection relates to one's own profile or account; what other people publish about someone may still be processed.

The same page states that the “Meta AI” chatbot, the blue circle in Messenger, Instagram and WhatsApp, cannot be switched off.

Source: Verbraucherzentrale, “Meta AI” bei Facebook, Instagram und WhatsApp (“so widersprechen Sie”) (in German), status 23 June 2025. Read on 18 October 2026. We found no newer Meta announcement of a new deadline for private users in the EU.

WhatsApp

The Verbraucherzentrale reports that, according to Meta (status 5 June 2025), interactions with AI on WhatsApp are not used to further improve Meta's AI, and that a separate objection form nevertheless exists, reached through WhatsApp settings and the privacy policy. It asks for the phone number of the WhatsApp account and an e-mail address. We did not find a statement by Meta on this that we could open ourselves on 18 October 2026; what is written here about WhatsApp is the Verbraucherzentrale's account, dated as above.

What MachineWitness adds

MachineWitness cannot object on anyone's behalf and does not replace the objection at Meta. What it offers is a free list, AI training: yes or no, on which a person records whether AI may be trained on their accounts and websites. The entry is kept with its date and sealed every night by two independent witnesses, so it can later be shown what was chosen and since when. The list obliges no AI company to anything, and a yes counts as much as a no.