If you use Facebook, Instagram, WhatsApp, or Threads, you’ll have seen Meta AI appear in your chats, search bar, or feed. Meta’s Llama models are designed to learn from data across Meta services, leaving many asking: how do I opt out of Meta AI?
In the UK and EU, data protection laws enable users to submit an objection request that limits how Meta uses their personal data for AI training purposes. However, the only way to completely stop AI data collection is to delete and stop using Meta services and devices.
Meta trains its AI using public posts, comments, photos, likes, Messenger conversations, and chatbot interactions. The level of data collection is alarming. User interactions are constantly analyzed to improve Meta’s generative AI systems, profile users, and refine ad targeting.
Meta claims that it does not use private messages to train its AI. However, if you invite Meta AI into a group chat, all participants’ messages are analyzed to improve Meta AI. Meta can also use AI data to make secondary inferences: combining user data to learn sensitive information such as sexual orientation, religious beliefs, or political preferences.
In this guide, I explain how to submit an objection request through Meta’s Privacy Center. This limits Meta’s AI training for users in select countries. I will also show you how to use privacy tools to reduce tracking and provide tips to help you control what you share with Meta AI.
Yes, but only in select regions. Users in the UK, EU/EEA, and Brazil can submit an objection request or refuse certain uses of their personal data for AI training. These settings are not available everywhere, and they won’t completely prevent Meta from using your data to train AI.
When you accept Meta’s Terms of Service on platforms like Facebook, Instagram, Threads, or WhatsApp, you allow Meta to use your data to improve its services, including Llama chatbot models. This is worrying because generative AI models may reproduce content from their training data in later conversations.
The possibility that large language models may accidentally leak personal information makes it crucial never to give a chatbot private medical details, your address, your date of birth, financial information, or other PII. It is also important to understand how much data Meta collects, and the limits of any opt-out requests.
Below, I have included the steps needed to limit Meta AI data collection. These settings are not available in all countries. If you’re unable to use these opt-out settings, continue using the rest of this guide to reduce the level of data collection and AI training you are exposed to.
Use the steps below to submit a Meta AI data collection objection request.
If you mainly use WhatsApp, Threads, Horizon, or a Meta device, start with the Facebook or Instagram account connected to your primary Meta Account email. If the form doesn’t appear, it may not be available in your country.
You can also try connecting to a VPN server in a country where the form is available. This may help you access the Meta AI objection form, but it is not guaranteed to work. Meta can also use your account region, phone number, location history, and other signals to determine whether you are a legitimate data subject in a country with stronger opt-out rights.
Want to opt out of Meta AI on Instagram instead? The steps are similar, but I have included them below so you can follow the process inside the Instagram app. If your Instagram and Facebook accounts are linked in Accounts Center, you may only need to submit the objection request once. Regardless of which route you take, wait for confirmation from Meta to find out whether your request was accepted.
In addition to objecting to Meta using information from your own Facebook or Instagram account, you may also be able to submit a request about personal information that Meta has obtained from third-party sources.
This is useful if Meta AI has produced, exposed, or appeared to know personal information about you that you suspect came from outside your account. I recommend using this feature if Meta AI displays your contact details, location, workplace, or any other Personally Identifiable Information (PII) in a response.
No. When you use services like Facebook, WhatsApp, Instagram, Horizon, and Threads, you must accept Meta’s Terms of Service. This gives Meta permission to collect and use personal data for profiling, targeted advertising, and service improvement.
This means the only way to prevent Meta from training its AI using your data is to delete your accounts. Even then, people you know could accidentally provide information about you to Meta if they talk about you, tag you, upload photos of you, or share your details in a message. This creates grey areas because Meta has been shown to collect and process personal data about people who have never signed up to Meta services, a practice often called shadow profiling.
Meta may also receive information from third-party websites, apps, advertisers, tracking pixels, embedded widgets, contact uploads, and other partner data sources. This creates a loophole because Meta may collect data about you from partners in its ecosystem via Meta Business Tools.
Studies have shown that Facebook can track users and non-users across the web, enabling it to infer personal information even from people who never signed up for Meta services. This means deleting your Meta accounts may not be enough to completely stop Meta from receiving or processing information about you.
The best course of action is to avoid sharing anything sensitive on Meta platforms. I also advise educating friends and family about how Meta uses data. This helps explain why it is important to protect not only your own information, but also the data of people in your circles.
If you live outside of the UK, the EU, or Brazil, you may decide to use a VPN to change your region. A VPN is an online privacy tool that hides your IP address from the websites you visit. By connecting to a VPN server in a country where Meta is required to process objection requests, you may be able to access the opt-out forms.
Note that you may also need to update your Meta settings to change your account’s home region. Even then, I can’t promise that using a VPN will work.
Meta reviews objection requests, and it can use your account region, usage history, phone number, location history, and account behavior to check whether you are genuinely a data subject in a country where opt-out rights are protected.
With this caveat out of the way, here are the steps to change your Meta region with a VPN:
If you still cannot find the AI usage objection form, you may also need to update your Facebook account region settings. I have provided those steps below.
Meta previously offered a setting that allowed some users to hide the Meta AI button in WhatsApp. This feature has since been removed, so you can no longer fully disable Meta AI from within WhatsApp. Instead, you need to avoid communicating with Meta’s chatbot from inside your WhatsApp account.
Meta states that private WhatsApp chats remain end-to-end encrypted and are not used for AI training. However, Meta AI conversations are different. If you message Meta AI directly or invite Meta AI into a group chat, the chatbot needs access to the relevant messages to respond. This means those messages are no longer private in the same way as ordinary WhatsApp conversations.
If you invite Meta AI into a group chat, all participants’ messages may be exposed to the chatbot. This can undermine the privacy people expect from WhatsApp because Meta AI can analyze the conversation and use that information to generate responses, improve its systems, or personalize Meta features.
You should also be wary when communicating with sellers or businesses on WhatsApp. Business accounts may use AI tools, customer service software, or Meta Business features that process messages differently from ordinary end-to-end encrypted personal chats. For this reason, I advise against sharing PII with businesses in WhatsApp messages.
Yes. Meta Quest headsets, Ray-Ban Meta glasses, and future Oakley Meta devices include cameras, microphones, voice controls, sensors, cloud features, diagnostics, and account-level safety systems. This creates elevated privacy risks.
Photos and videos shared with Meta AI from smart glasses may be used to improve Meta AI, and AI-processed photos may be used to train Meta’s AI. This is concerning because smart glasses can capture bystanders who never accepted Meta’s terms, potentially contributing to shadow profiles.
Quest headsets raise similar concerns because they rely on cameras, microphones, movement data, hand tracking, screenshots, cloud features, diagnostics, age checks, and account systems. Research shows that VR sensor data can identify users, and Quest users have expressed concerns that Meta uses this capability to detect when a minor is using the device.
To maximise your privacy, disable cloud backups where possible, review camera and microphone permissions, avoid voice commands, turn off diagnostics, avoid Meta AI features, delete stored voice recordings, and power down the device when it is not in use. Also consider a VPN for Quest.
Users who want to avoid hardware-level collection should delete their Meta account and stop using Meta hardware.
Meta originally trained its Llama models using large datasets, licensed content, and publicly available information from the internet. As public training data becomes harder to source, Meta has moved on to harvesting data from its own ecosystem of services and devices: Facebook, Instagram, WhatsApp, Threads, Ray-Ban Meta glasses, Quest headsets, and other Meta services. This gives the company access to a steady stream of fresh training data.
Meta uses public posts, photos, captions, comments, likes, chatbot interactions, and device-based data. Anything you post publicly, share with Meta AI, or provide during an AI chatbot conversation or group chat could be processed to train Meta AI.
The California Consumer Privacy Act (CCPA) gives Californians rights over their personal information, including access, deletion, correction, and opt-outs for some data sharing. Unfortunately, this is not the same as the GDPR right to object to Meta using your data for AI training.
Users in California may not be able to access the Meta AI objection form. Instead, you will need to tighten your privacy settings, delete public posts you don’t want used to train Meta AI, avoid Meta AI inside your accounts, and use CCPA controls to limit sale, sharing, or certain uses of sensitive personal information where available.
You can’t use privacy apps and extensions to stop Meta from collecting publicly available data or interactions you make with its services. This is because you are directly logged into your Meta device or account, which means Meta has access to your posts, interactions, and usage telemetry.
That said, tools like Norton AntiTrack, uBlock Origin, Privacy Badger, Ghostery, Brave, Firefox, DuckDuckGo Privacy Essentials, and a VPN can help limit third-party tracking carried out by Meta business partners and affiliates.
These tools disguise your digital fingerprint and block tracking scripts that Meta uses to follow you around the web for profiling and advertising. A VPN also hides your IP address from websites and services that use Meta Business Tools, helping to reduce one important layer of off-platform tracking.
Some deadlines have already passed, and opting out now may not undo training that already happened using previous posts and interactions. However, it is still worth checking whether an opt-out form is available in your region. If it is, submitting an objection request can help limit how Meta uses your data for future AI training.
Many users say they can’t find the Meta AI opt-out form. In most cases, this is because the opt-out is not available in the country where they live. That said, finding the opt-out form can be hard even if you live in the UK, EU, Brazil, or another region where Meta provides an objection route. I have provided the steps to find and submit the opt-out form in this guide.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。