1. What the document is: "Meta AIs Terms of Service (EU)" — the EU-region-specific version of Meta's consumer generative-AI terms, governed by Meta Platforms Ireland Limited. The document explicitly distinguishes itself from separate rest-of-world, UK, and Brazil versions, which are referenced but not captured in this dataset.
What this wiki found — complete, every page
Presence facts, not verdicts — each flagged term links to its definition and the exact clause on this page. Absence of a badge means the term isn't currently flagged here, not that the page is risk-free.
Independent reviews, where they had something to say
Meta AI – Terms of Service
Dataset: GenGA (Generative AI Governance Archive) — 11 AI services, 2025–present
GenGA (Generative AI Governance Archive) is this wiki’s Generative-AI dataset: raw policy snapshots captured by the Open Terms Archive
genai-euproject across 11 GenAI providers (ChatGPT, Claude.ai, DeepSeek, Google Generative AI Services, Le Chat, Llama API, Meta AI, Microsoft Copilot, Perplexity, Qwen Chat, xAI). Unlike PGAv2, GenGA has no pre-tagged risk scores — all risk scoring and keyword tagging on these pages is LLM-assigned by direct reading, spanning 2025–present.
Note: This page contains documented policy clauses. Risk assessment is qualitative and context-dependent. For analysis of patterns across platforms, see:
Source: sources/GenGA/Meta AI/Terms of Service/ (raw Markdown captures, Open Terms Archive genai-eu dataset — no pre-tagged JSONL exists for this source)
Last updated: 2026-01-21 (only capture in this dataset)
Note on methodology: No pre-computed
risk_score/keywordfields exist for this source. Keyword tags and clause analysis on this page are LLM-assigned, using the same rubric/vocabulary as the PGAv2 pages plus new GenAI-specific tags (input license,feedback). Treat as first-pass analysis, not externally verified ground truth. This wiki’s risk-scoring system was retired project-wide on 2026-06-21 (seemethodology.md§4) — no page, GenGA included, computes or displays a numeric risk score.
Overview
1. What the document is: “Meta AIs Terms of Service (EU)” — the EU-region-specific version of Meta’s consumer generative-AI terms, governed by Meta Platforms Ireland Limited. The document explicitly distinguishes itself from separate rest-of-world, UK, and Brazil versions, which are referenced but not captured in this dataset.
2. Input rights: You “retain all right, title, and interest” in your Prompts. But Meta’s usage rights over “Content” — a bundled term covering Prompts, Outputs, and Feedback together — are described only in broad language: “to provide, maintain, and improve Meta services and features” and “to conduct and support research.” The actual training-data mechanics are deferred to the separate Meta Privacy Policy (“explains… how we use your interactions with AIs to improve AI at Meta”), which is not captured in this dataset, and this document itself describes no opt-out.
3. Output restrictions: Standard accuracy/reliability disclaimers only — no licensing restriction on how you use Outputs.
4. Non-explicit predatory clauses: Major finding: the “Acceptable Uses” list — what users may not do with the AIs — closely tracks the EU AI Act’s Article 5 list of prohibited AI practices. It bans subliminal or manipulative behavioral distortion, exploiting age/disability/economic vulnerability, social-behavior scoring that leads to “detrimental or unfavorable treatment,” predictive policing and criminal-risk profiling, untargeted facial-image scraping for recognition databases, workplace/education emotion recognition, categorizing people by biometric traits, and real-time public-space biometric identification for law enforcement. Yet the document never cites the AI Act, or any regulation, by name. This is the clearest case in the dataset of AI-Act-derived substance adopted without AI-Act attribution — the inverse of Le Chat’s explicit AI Act citation (which appeared briefly, then was removed).
5. Regulatory references: None found by name. No GDPR or AI Act citation appears, despite this being the EU-specific version of the document and substantively mirroring AI Act Article 5 content (see above).
6. Regional variation: The top of the document explicitly splits out rest-of-world, UK, Brazil, and this EU version — the clearest, most explicit regional segmentation statement in this dataset.
7. Key risk to users: Moderate. The main risk is the vague “improve services”/research usage grant over Content, which has no opt-out. That’s balanced by genuinely user-protective features: “Private Processing” technology, under which “Meta cannot read or access” private messages shared with AI features that use it, plus the AI-Act-aligned behavioral protections.
Flagged Keywords & Risks (LLM-assigned)
improve services— Meta can use all your Content (Prompts, Outputs, and Feedback) “to provide, maintain, and improve Meta services and features” and “to conduct and support research,” with no opt-out stated in this document. Why it matters: this is the vaguest, least-specific Content-usage clause in the dataset — no tier exceptions, no clear training/non-training line, and the real mechanics are deferred entirely to a Privacy Policy this wiki hasn’t captured.feedback— Feedback is folded into the single “Content” definition alongside Prompts and Outputs, rather than getting its own separate clause. Why it matters: this is broader and less precisely scoped than other providers’ dedicated Feedback clauses — you can’t easily tell which usage rights apply specifically to Feedback versus Prompts/Outputs.
Regulatory & Research Context
Meta AI was not among the providers studied by Davidson et al. (2026) or Pandit et al. (2026), but their frameworks apply by analogy: Pandit et al.’s finding that GenAI terms broadly “discard assurances” while leaving users to shoulder responsibilities they cannot materially fulfil resonates with the improve services clause’s vague, un-opted-out research/improvement grant over Content, deferred to an uncaptured Privacy Policy. The page’s own “Major finding” — that Meta’s Acceptable Uses list substantively mirrors the EU AI Act’s Article 5 prohibited-practices list without ever citing the Act by name — is precisely the kind of unattributed regulatory borrowing that Edwards et al. (2025) frame within the “platformisation paradigm,” where providers position themselves as neutral intermediaries while quietly retaining control over how (and whether) such obligations are disclosed.
Changes Summary
| Date | What changed |
|---|---|
| 2026-01-21 | Baseline and only version captured in this dataset (“Effective Date: January 5, 2026”). |
Version History
2026-01-21
- Explicit AI clause: YES
- Non-explicit predatory: YES
- Flagged keywords:
improve services,feedback
Clause: improve services
Meta uses Content and related information, including through automated or manual (i.e. human) review and through third-party vendors in some instances, including: To provide, maintain, and improve Meta services and features. To conduct and support research. […]
Clause: AI Act-aligned Acceptable Uses (not separately concept-tagged — no AI Act citation present)
You may not access, use, or allow others to access or use AIs in any manner that would: […] deploy subliminal, manipulative, or deceptive techniques to distort behavior by impairing decision making; or exploit people due to their age, disability, or a specific social or economic situation to distort their behavior. 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.
Clause: Private Processing (not separately concept-tagged — user-protective)
When you share your private messages with friends and family with AIs that use Private Processing technology, Meta cannot read or access the messages you have shared. Meta provides Private Processing as a service on your behalf.
Clause: feedback
AIs allows you to provide materials for processing… and will generate a response (“Outputs”) based on your Prompts, as well as any feedback you provide (“Feedback”) collectively, “Content.”