Free exposure scan. Cleanup plans start at $1 for the first month. Start cleanup →
Home/Blog/Meta AI Training Your Photos
Back to Blog
Privacy Threat

Meta Trains AI on Public Posts and Photos: What to Do

9 min read

Meta says it uses public content shared by adults on its products to train its AI. If your Facebook or Instagram posts, photos or comments are public, assume they can be used. European users can object through a form. In the sources we found, US users have no comparable form.

What Meta Says It Uses for AI Training

Meta describes how it uses information for generative AI on its Privacy Center. In its April 14, 2025 announcement for Europe, Meta said it trains its AI on public content shared by adults on its products, does not use public data from EU accounts of people under 18, and does not use the content of private messages with friends and family unless you or someone in the chat chooses to share those messages with its AIs.

That announcement was about the EU. We did not find an equivalent statement written for US users, so read the Privacy Center page for what applies to you.

What "Public" Usually Means

  • Any post with its audience set to "Public"
  • Any comment on a public Page, public Group, or public post, even if your own profile is private
  • Any Instagram post on a non-private account

The EU vs. US Divide

European Union: The Right to Object

Meta paused its plans to train on EU user data in June 2024 after the Irish Data Protection Commission, its lead EU regulator, raised concerns. In April 2025, Meta announced it would resume, with notices to EU users that include a link to a form to object "at any time." In a May 21, 2025 statement, the Irish regulator said Meta had provided updated documentation for training to begin on May 27, 2025, and that it had recommended changes that Meta implemented. The right to object comes from Article 21 of the GDPR, which applies to processing based on legitimate interests.

United States: No Comparable Form in the Sources We Found

MIT Technology Review reported in June 2024 that Meta had no opt-out feature for people in places without national data privacy laws, and suggested setting accounts to private to reduce risk. We did not find a US objection form as of November 2026. Meta can change this, so check its Privacy Center.

The United States has no comprehensive federal privacy law. California's CCPA gives residents the right to opt out of the sale or sharing of personal information. Whether Meta's in-house AI training counts as either is a separate legal question that this post does not settle.

Meta AI and Cloud Processing: Your Camera Roll

In 2025, Facebook began offering an opt-in camera roll "cloud processing" feature. Malwarebytes reported on July 1, 2025 that it is opt-in, and that Meta's AI terms allow analysis of uploaded images "including facial features." The same report says users should not upload images of people who live in Illinois or Texas unless they are legally authorized to consent for them. We could not read Meta's own help page, so check your settings and Meta's current terms for what is retained and whether it is used for training.

Illinois and Texas Biometric Laws

Illinois's Biometric Information Privacy Act lets a person recover $1,000 for a negligent violation or $5,000 for an intentional or reckless one, or actual damages if greater, plus attorneys' fees. Texas's biometric identifier law (Bus. & Com. Code section 503.001) covers capturing biometric identifiers for a commercial purpose. We are not saying cloud processing violates either law, and we have not seen Meta explain the Illinois and Texas wording.

The Tagging Problem: Your Face in Someone Else's Photo

Your privacy settings control your own posts. They do not control a public photo someone else posts of you. If a friend posts a public photo of your birthday dinner and tags you, that photo is public content, and your face, name tag, location and date are visible in it.

The same applies to workplace photos on an employer's public page, event photos posted by public groups, and any other public content where you appear. If Meta trains on public content, anyone who appears in it is included whether or not they posted it.

What You Can Actually Do

We found no US objection form, but you can reduce how much public content exists. Menu names change, so look for the equivalent setting if a label below has moved.

  1. Set your Facebook posts to friends only. Change "Who can see your future posts?" and use the "Limit Past Posts" tool for older public posts. This does not undo anything already collected.
  2. Switch your Instagram account to private. Settings, then Account Privacy, then "Private Account." Meta says it trains on public content. Posts made while your account was public may already have been collected.
  3. Review and remove tags from public posts, and ask friends to make photos of you non-public.
  4. Decline camera roll cloud processing if Facebook offers it.
  5. Be deliberate about public comments. Comments on public pages, groups and posts are public content regardless of your profile's privacy settings.
  6. Download your data with Meta's "Download Your Information" tool to see what Meta holds. This will not stop AI training.

The Data Broker Connection

Meta's AI training and data brokers are separate. Reducing your public social media footprint limits what is visible on Meta. Removing your data from broker databases targets a different source of exposure. See our social media privacy guide. GhostVault files deletion demands with data brokers on your behalf, from $69 a year or $7.99 a month. We cannot guarantee that any broker deletes your data.

The Regulatory Outlook

The American Privacy Rights Act, which would have been a comprehensive federal privacy law, was introduced in 2024 and did not pass. Until a law changes that, the practical protection for US users is to minimize the amount of public content available for collection.

Frequently Asked Questions

Is Meta using my photos to train AI?

Meta says it uses public content shared by adults on its products to train its AI. For the EU, it stated in April 2025 that it does not use the content of private messages with friends and family unless you or someone in the chat chooses to share them with its AIs. If your posts are public, assume they can be used. Comments on public pages and posts are public even if your profile is private.

Can I opt out of Meta using my data for AI training in the US?

In the sources we found, no. MIT Technology Review reported in June 2024 that Meta had no opt-out for people in places without national data privacy laws, and we found no comparable US form as of November 2026. EU users can object through a form. Check Meta's Privacy Center for current options. The most effective step for US users is to make profiles private and limit public interactions.

Does Meta use private messages to train AI?

Meta has said it does not use the content of private messages with friends and family to train its AIs, unless you or someone in the chat chooses to share them with its AIs. Conversations you have directly with Meta AI are different from messages between people, so assume they can be used.

What happens if someone tags me in a photo that Meta uses for AI training?

If someone posts a public photo of you, it is public content and your own privacy settings do not control it. To address it, untag yourself, ask the poster to change the photo's audience to non-public, or ask them to remove it.

How does Meta AI training relate to data brokers?

They are separate issues. Reducing your public social media presence limits what is visible on Meta. Removing your data from broker databases, with a service like GhostVault, targets a different source of exposure. We cannot guarantee that any broker deletes your data.

Related guides

Popular on GhostVault