Does Meta train on your data?
Meta's UK policy names developing and improving AI technology as a purpose and public posts as its input. The way out is an objection form on Meta's own AI page, not in the policy.
"In our interest, and in the interests of users and third parties, to develop and improve artificial intelligence technology while ensuring appropriate safeguards, such as improving model responses for safety and accuracy."
- 6 Sept 2026
- UK accounts
- uk
- Saved by hand — not re-read automatically
- F5A62971…55D9
How we read it: Graded on the UK-served text, saved by hand on 6 September 2026. Its legitimate-interests table names developing and improving artificial intelligence technology as a purpose and lists the information used for it: public content you or others create or share, and your interactions with the features that are part of AI at Meta. No form of "train" appears; this site reads "develop and improve" AI on named inputs as it read Booking's and Vinted's. The policy itself names no route. Meta's separate page on generative AI, saved the same day, does: an objection request through your account, confirmed by email. A form to go looking for is a D, and the steps quote that page. Without it this was an F.
opt out, object, right to object, withdraw, settings, legitimate interest, AI at Meta
What the document does offer
- "legitimate interest" appears once, in the personalisation purpose for users under the age of majority; the AI-development purpose is stated as an interest without naming a right to object against it
- "AI at Meta" appears five times: as a feature Meta provides, and as an information category it collects ("Interactions with our features that are part of AI at Meta and related metadata")
- "opt out", "object", "right to object" and "withdraw": zero times each
Why none of it applies: No control over AI development is named in the UK-served text: not a setting, not a form, not an objection route. The right to object that legitimate-interests processing carries under UK law is not mentioned, so a reader of this document is not told it exists.
The US-served text, captured by the Internet Archive on 1 June 2026, lists interactions with AI at Meta among the information collected and names no AI-development purpose at all: no form of "train" and no "model" anywhere, and the only "artificial intelligence" is a sentence about supporting research. The UK-served text of 6 September 2026 that grades the tier above is a different document, and whether the US policy now says the same is not something this site can read from here: facebook.com refuses this monitor in robots.txt, and the Internet Archive copies since June arrived in Russian. Unresolved, rather than inferred from the UK text.
- Meta Privacy Policy, US-served text, Internet Archive capture of 2026-06-01 (the 2026-07-01 and 2026-08-01 captures were served in Russian; the UK-served text saved by hand on 2026-09-06 is a different document and grades the UK tier)
- train, training, trains, model, models, artificial intelligence, machine learning
- 0 of 2 raw hits
- 13 Aug 2026
"Train", "training", "trains", "model" and "models" appear zero times each. The two matches are a single "artificial intelligence" and a single "machine learning", both in one sentence — "We support research in areas like artificial intelligence and machine learning" — which describes research funding, not a use of your content.
What each plan can refuse
The same questions, asked of every plan we could identify. Every cell comes from a quoted clause or a documented search - see the evidence above.
- Contact support
- Not stated
- 56 / 100
- Not stated
- Unknown
- Not established
- Not stated
- —
- —
- Not established
How to opt out of Meta's AI training
Plan by plan, from what Meta's own policy says about refusing. Where it publishes the steps, they're listed here with the limits they carry. Where there's nothing to switch off, or nothing graded, this says so instead of inventing a menu path.
UK accounts
Yes, but not self-serve: Meta makes you contact support.
- Log in to the account you use with AI at Meta.
- Open facebook.com/privacy/genai/ and, under "You have rights related to how your information is used for AI at Meta", follow the link for your account to the objection form.
- Submit the objection. Meta emails a confirmation; an objection made after a training run has started applies to future runs.
US accounts
Not graded, so no route is pointed at. The US-served text, captured by the Internet Archive on 1 June 2026, lists interactions with AI at Meta among the information collected and names no AI-development purpose at all: no form of "train" and no "model" anywhere, and the only "artificial intelligence" is a sentence about supporting research. The UK-served text of 6 September 2026 that grades the tier above is a different document, and whether the US policy now says the same is not something this site can read from here: facebook.com refuses this monitor in robots.txt, and the Internet Archive copies since June arrived in Russian. Unresolved, rather than inferred from the UK text.
What this policy has done since we started watching
5 changes detected since 1 Jan 2023. Each one is a diff between two archived copies, and each links to the passage that moved.
- Touched opt-outs, retention, what's collected, sub-processors and training.
- Touched opt-outs, retention, what's collected, sub-processors and training.
- Touched what's collected, sub-processors and training.
- Touched what's collected. Not yet reviewed by a human.
- Touched opt-outs, retention, what's collected, sub-processors and training.
Everything we read
Vendor response
We wrote to Meta (Facebook / Instagram) on 6 Sept 2026, quoting the clauses above and inviting a correction. Nothing has come back. That's published because a company declining to comment on its own policy is something a reader should know, and because the invitation should be on the record whether or not it's taken up. Any reply goes here, word for word.
Report a wrong clause