A Big One

Meta's global Meta AI Terms of Service shrank this week. The list of prohibited user activities went from 14 bullets to 8. The six deletions are not random.

What changed: Six prohibitions, all tracking EU AI Act Article 5 language, were removed from the global Meta AI ToS:

• "Evaluate or classify individuals based on social behavior or personal traits that leads to detrimental or unfavorable treatment..." • "Engage in predictive policing or criminal risk assessments based on profiling or assessing personality traits and characteristics" • "Engage in the untargeted scraping of facial images from the internet or CCTV footage to build facial recognition databases" • "Engage in emotion recognition in the workplace or education, save as permitted by applicable law" • "Categorise people based on their biometric data to deduce or infer their race, political opinions, trade union membership, religious or philosophical beliefs, sex life or sexual orientation" • "Engage in real-time biometric identification in publicly accessible spaces for the purposes of law enforcement, save as permitted by applicable law"

The regional pointer block — a passage in the prior ToS directing users to region-specific versions in the UK, Brazil, and the European Region — was also removed from the global document. The prior referenced three separate AI terms documents at /legal/uk-ai-terms, /legal/br-ai-terms, and /legal/eu-ai-terms. The Ledger does not currently track any of those three.

Two smaller changes round out the update. The word "psychological" was dropped from the list of professional-advice categories users are prohibited from soliciting (medical, financial, and legal remain). And the document was reformatted from a bold-header structure to a numbered-section structure that reads more like a formal contract.

In the same scrape window, Meta's Privacy Policy – Supplemental had AI-specific language removed from its Voice Interactions section — references to "Meta AI," "generative AI powered non-player characters," and "AI-powered features" were replaced with generic phrasing. The supplemental policy's effective date was bumped to March 5, 2026.

What we don't yet know: Whether the six deleted prohibitions were preserved in Meta's regional AI terms documents, or whether they were dropped from those documents too. The Ledger is adding the EU Meta AI ToS to daily tracking starting this week, and the UK and Brazil versions will follow.

Why this might matter for AI content producers: The six removed clauses all map to EU AI Act Article 5 — the article that defines prohibited AI practices. If you're producing AI content using Meta AI for non-EU markets, the open question is whether the deletion means those activities are no longer contractually prohibited globally, or whether they remain prohibited through regional documents incorporated by reference. Either is possible from what's visible in the global ToS alone. The pattern is also worth watching: this is the second platform in three weeks to split a global policy along regional lines (see Vol. 26 No. 10 on OpenAI's privacy policy). The shape of that split — whether the global document loses protections the regional versions preserve, or whether both shift together — is something we'll be able to assess once the EU document is in the corpus.

In human terms: A compliance officer at a marketing agency previously included Meta's six Article 5-aligned prohibitions in the agency's internal AI-use checklist (the agency cited Meta's own ToS rather than writing their own rules). After this week's deletion, those citations point to language that no longer exists in the document — the officer either has to track down regional AI terms documents that may carry equivalent language, or rewrite the prohibitions in the agency's own policy. The work the global ToS used to do for them now falls to them.