Skip to main content
Halo Aware
Supervision advised53/100for under-18 use

Is Meta AI safe for kids?

Under our published methodology, Meta AI scores mid-cohort for under-18 use: it now ships teen-specific protections and parent-facing topic insights, but these followed the August 2025 Reuters-reported internal standards that had permitted 'romantic' and 'sensual' chats with minors, an FTC 6(b) order and a 44-state AG letter. Meta's terms state AI interactions are used to personalise ads and improve AI, with no opt-out described, and deleting chats 'may not delete our copy'.

Checked 29 July 2026 · An assessment by Halo Aware under our published methodology

How it scored

Each dimension is scored 0–4 against published anchors and must cite the evidence below. The overall score is weighted arithmetic, not judgement calls.

DimensionScoreEvidence
Content safety×252/4e4, e6, e7, e10, e11Meta publishes teen-specific standards: PG-13-inspired responses, AI characters designed not to discuss self-harm, suicide or disordered eating with teens, signposting to expert resources (e6, e7); terms bar self-harm content (e4). Held at 2 given the documented 2025 failures (e10, e11).
Age controls×203/4e1, e6, e7The AI Terms set a 13+ minimum (e1), and Meta states it uses AI technology to place suspected teens into Teen Account protections even if they claim to be adults, alongside a distinct teen AI experience (e6, e7). Short of 4: the age detection's effectiveness is not independently verified.
Data practices×201/4e2, e3, e5The terms state AI interactions personalise experiences and ads and improve AI at Meta, and that AIs sometimes retain shared information (e2, e5); deleting messages or an account 'may not delete our copy' (e3). Deletion paths exist via the Privacy Center; no training opt-out is described (e5).
Design pressure×152/4e6, e7, e10, e11Meta AI is a utility-framed assistant, but it shares its surface with companion-style AI characters, and reported pre-remediation standards allowed romantic roleplay with minors (e10, e11). Scored 2: assistant framing, offset by character mechanics now restricted for teens to a limited set (e7).
Parental visibility×103/4e6, e8, e12Shipped supervision: an Insights tab shows topics a teen asked Meta AI about in the past week (e8); announced controls turn off one-on-one AI character chats and block characters (e6). Not a 4: controls cover AI characters only, no core-assistant control described (e6); insights region-gated (e8).
Track record×102/4e9, e10, e11, e6, e8Reported internal standards permitted 'romantic' and 'sensual' chats with minors (e10, e11); FTC 6(b) orders named Meta Platforms and Instagram (e9). Offset by dated, shipped remediation (e10, e6, e8); no litigation over harm to a minor by Meta AI was found. Scored 2.

What we found

Every claim is sourced and dated. If something here is wrong or out of date, tell us and we will fix it.

  1. e1Meta AI Terms of Service (effective 13 May 2026) state users cannot access Meta's AIs if under age 13, or a greater age where required in their country (read via Internet Archive snapshot dated 2026-06-29; the live page blocks automated access; checked 2026-07-29).

    "Without limiting the Meta Terms of Service or any other applicable terms or policies, you cannot access AIs if you are under the age of 13 (or such greater age required in your country or territory) or if you are prohibited from receiving our services or software under applicable laws."

    https://web.archive.org/web/20260629061903/https://www.facebook.com/legal/ai-terms · captured 29 July 2026

  2. e2The Meta AI Terms state interactions with AIs are used to personalise experiences and ads and to 'improve AI at Meta', that AIs 'will sometimes retain and use' shared information, and warn users not to share information they do not want retained (via Internet Archive snapshot, checked 2026-07-29).

    "The Meta Privacy Policy explains how we process personal information in the context of AIs, including how we use your interactions with AIs to personalize your experiences and ads, and improve AI at Meta. […] When information is shared with AIs, the AIs will sometimes retain and use that information. Do not share information that you don’t want the AIs to use and retain, such as information about sensitive topics."

    https://web.archive.org/web/20260629061903/https://www.facebook.com/legal/ai-terms · captured 29 July 2026

  3. e3The Meta AI Terms state that deleting individual messages, entire threads, or even an account 'may not delete our copy of your personal information', directing users to Meta's Privacy Center for deletion (via Internet Archive snapshot, checked 2026-07-29).

    "It is important that you review the Privacy Center before you try to delete any personal information or delete or deactivate your account. Depending on how you use AIs, deleting individual messages, entire threads, or even your account may not delete our copy of your personal information, and the Privacy Center describes the correct way to delete your personal information."

    https://web.archive.org/web/20260629061903/https://www.facebook.com/legal/ai-terms · captured 29 July 2026

  4. e4The Meta AI Terms' acceptable-use rules prohibit using AIs in any manner that would generate or facilitate content that may harm individuals, including encouraging self-harm, and any form of child sexual abuse material (via Internet Archive snapshot, checked 2026-07-29).

    "You may not access, use, or allow others to access or use AIs in any manner that would: Violate any applicable law (including but not limited to any manner that promotes, contributes to, or encourages illegal or unlawful activity), including conduct that describes, encourages, supports, or distributes in any form child sexual abuse materials Generate, promote, disseminate, or otherwise use or facilitate content that may harm other individuals, including to encourage self-harm, discrimination, bullying, humiliation, and/or other emotional harms."

    https://web.archive.org/web/20260629061903/https://www.facebook.com/legal/ai-terms · captured 29 July 2026

  5. e5Meta's Privacy Policy lists 'Interactions with AI at Meta and related metadata', including content and messages exchanged with the AI, among information used to 'Develop and improve AI at Meta for Meta Products and for third parties' (via Internet Archive snapshot dated 2026-07-02, checked 2026-07-29).

    "Developing and improving AI at Meta Why and how we process your information: Develop and improve AI at Meta for Meta Products and for third parties. […] Interactions with AI at Meta and related metadata. For example, information you or others exchange with AI at Meta like content and messages."

    https://web.archive.org/web/20260702022542/https://www.facebook.com/privacy/policy/ · captured 29 July 2026

  6. e6Meta announced (newsroom, 17 Oct 2025) parental controls letting parents turn off teens' one-on-one chats with AI characters and block specific characters, and stated it uses AI technology to place suspected teens into Teen Account protections even if they claim to be adults.

    "We’re introducing simple, new ways for parents to turn off their teens’ access to one-on-one chats with AI characters, and giving them more insights on how teens are interacting with AI. These new parental controls are in addition to the protections we already automatically apply to Teen Accounts, and we use AI technology to place suspected teens into these protections, even if they tell us they’re adults. […] Parents will be able to block specific AI characters if they don’t want to turn off access to AI characters altogether."

    https://about.fb.com/news/2025/10/teen-ai-safety-approach/ · captured 29 July 2026

  7. e7Meta states (newsroom, 17 Oct 2025) that teen AI experiences follow an age-appropriate, movie-rating-inspired standard and that AI characters are designed not to engage teens in age-inappropriate discussions of self-harm, suicide or disordered eating, directing teens to expert resources when appropriate.

    "we updated AI experiences for teens to be inspired by movie ratings for ages 13+ and parent feedback , meaning our goal is that AIs should not give age-inappropriate responses that would feel out of place in an age-appropriate movie. […] AI characters are designed to not engage in age-inappropriate discussions about self-harm, suicide, or disordered eating with teens, or conversations that encourage, promote, or enable these topics. Our AIs are designed to respond safely to these topics and direct teens, when appropriate, to expert resources or support."

    https://about.fb.com/news/2025/10/teen-ai-safety-approach/ · captured 29 July 2026

  8. e8Meta announced (newsroom, 23 Apr 2026) that parents supervising Teen Accounts in the US, UK, Australia, Canada and Brazil can now see an Insights tab showing the topics their teen asked Meta AI about in the past week, with rollout to supervising parents globally to follow.

    "These insights are now available for parents supervising Teen Accounts in the US, UK, Australia, Canada, and Brazil, and they’ll roll out to supervising parents globally in the coming weeks. […] Parents using supervision on Facebook, Messenger, or Instagram will now see a new Insights tab within supervision, both in-app and on web. From there, parents will be able to see the topics their teen has been asking Meta AI about in that specific app over the past week."

    https://about.fb.com/news/2026/04/helping-parents-understand-conversations-their-teens-are-having-with-ai/ · captured 29 July 2026

  9. e9Documented recordThe US FTC issued 6(b) orders on 11 Sept 2025 to seven consumer AI chatbot operators, including Meta Platforms, Inc. and Instagram, LLC, seeking information on how they measure, test and monitor potentially negative impacts of chatbots on children and teens.

    "The Federal Trade Commission is issuing orders to seven companies that provide consumer-facing AI-powered chatbots seeking information on how these firms measure, test, and monitor potentially negative impacts of this technology on children and teens. […] The recipients include: Alphabet, Inc.; Character Technologies, Inc.; Instagram, LLC; Meta Platforms, Inc.; OpenAI OpCo, LLC; Snap, Inc.; and X.AI Corp."

    https://www.ftc.gov/news-events/news/press-releases/2025/09/ftc-launches-inquiry-ai-chatbots-acting-companions · captured 29 July 2026

  10. e10Documented recordCNBC reported (29 Aug 2025) that Meta made interim changes after a Reuters report that internal policy had permitted its AI chatbots to engage in 'romantic' and 'sensual' conversations with teens and children: retraining chatbots not to respond to teens on self-harm, suicide or disordered eating, limiting teens to educational AI characters; a US Senate probe was opened.

    "The social media giant is now training its AI chatbots so that they do not generate responses to teenagers about subjects like self-harm, suicide, disordered eating and avoid potentially inappropriate romantic conversations, a Meta spokesperson confirmed. […] Last week, Sen. Josh Hawley , R-Mo., said that he was launching an investigation into Meta following a Reuters report about the company permitting its AI chatbots to engage in "romantic" and "sensual" conversations with teens and children."

    https://www.cnbc.com/2025/08/29/meta-ai-chatbot-teen-senate-probe.html · captured 29 July 2026

  11. e11Documented recordA bipartisan letter from 44 US state attorneys general (25 Aug 2025) to AI companies including Meta stated that internal Meta documents revealed the company had authorised AI assistants to 'flirt and engage in romantic roleplay with children' as young as 8 (per the Illinois Attorney General's release).

    "Attorney General Kwame Raoul today co-led a bipartisan coalition of 44 attorneys general warning major artificial intelligence (AI) companies to stop hurting kids. The letter , sent to Anthropic, Apple, Chai AI, Google, Luka Inc., Meta, Microsoft, Nomi AI, Open AI, Perplexity AI, Replika and xAI addresses alarming reports of AI chatbots engaging in sexually inappropriate conversations with kids. Internal Meta documents reveal that the company authorized its AI assistants to “flirt and engage in romantic roleplay with children” as young as 8 years old."

    https://illinoisattorneygeneral.gov/news/story/attorney-general-raoul-leads-44-states-in-demanding-companies-end-predatory-ai-interactions-with-kids · captured 29 July 2026

  12. e12Observed by Halo AwareThe web surface at www.meta.ai refused our automated requests (HTTP 403) during this check, so we could not review the chat surface or any logged-out experience first-hand (checked 2026-07-29).

    https://www.meta.ai · captured 29 July 2026

Platform operator, or spotted an error? Email alerts@haloaware.com. We review disputed claims promptly, correct what is wrong, and log corrections publicly on this page.

Halo Aware doesn't monitor Meta AI yet

We add platforms parents ask about. Tell us Meta AI matters to your family and it moves up the roster.

Tell us it matters