Is Meta AI safe? Teen safeguards, privacy and the investigations

The short answer

It depends on what you share and who is using it. As of October 21, 2026, Meta says teens can use the Meta AI assistant with age-appropriate protections, but their access to AI characters has been paused since early 2026. Parents who supervise a teen account can see the topics their teen asks Meta AI about, not the conversations.

On privacy, Meta says it uses people's interactions with its AI to personalize content and ads, and that both automated systems and people review AI interactions. Officials have raised concerns about Meta's chatbots and children; those are allegations and inquiries, not findings.

Where Meta AI and AI characters appear

Meta AI is an assistant inside Facebook, Messenger, Instagram and WhatsApp, and in the standalone Meta AI app, which Meta launched on April 29, 2025.

AI characters are separate personas. Since July 2024, Meta's AI Studio let people create AI characters and share them on Instagram, Messenger, WhatsApp and the web. On August 10, 2026, Meta said people can no longer create new AI characters or edit existing ones in AI Studio; existing characters and chats remain active.

Who can use it and what teens get

Meta says it uses AI to place people it suspects are teens into teen protections, even if they say they are adults, and that it is investing in technology to catch accounts that may belong to children under 13.

• Teen access to AI characters is paused. On January 23, 2026, Meta said it would pause teens' access to existing AI characters globally while it builds a new version. The pause covers accounts with teen birthdays and suspected teens found through age prediction.
• Teens can still use the Meta AI assistant, which Meta says has default age-appropriate protections.
• Meta says its AIs should not give teens responses that would feel out of place in a movie rated for ages 13 and up, and that Meta AI is trained not to have sexual or romantic conversations with teens.
• In October 2025, Meta said its AI characters were designed to avoid age-inappropriate discussions of self-harm, suicide or disordered eating with teens and to point them to expert resources.
• Parents can switch the Meta AI app from its default 13+ setting to Limited Content, which Meta says further restricts teens' conversations.

Parental controls: announced vs. launched

On October 17, 2025, Meta announced parental controls for teens' AI use: turning off one-on-one chats with AI characters, blocking specific characters, and insights into the topics teens discuss with AI. It said they would start on Instagram early in 2026. After the January 2026 pause, Meta said the character controls would apply to the new version of AI characters. We did not find a Meta announcement that the character controls have launched.

What Meta says has launched:

• Insights tab (April 23, 2026): parents who supervise a teen account on Facebook, Messenger or Instagram can see the topics their teen asked Meta AI about over the past week, such as school or health. Meta says parents see topics but not the conversations. Meta said the feature launched first in the US, UK, Australia, Canada and Brazil. On October 5, 2026, Meta said supervision was available globally in the Meta AI app.
• Distress alerts (July 16, 2026): Meta said parents using Instagram supervision in the U.S., UK, Australia and Canada are alerted if a teen discusses suicide or self-harm with Meta AI, with a global rollout planned by the end of 2026. Meta says flagged chats are reviewed by a person before an alert is sent.
• Meta also said in July 2026 that it is building ways to contact emergency services for people who may be at imminent risk. That feature had not launched as of that post.

How Meta uses your AI chats

• Ads and recommendations: Meta announced on October 1, 2025 that it would use people's interactions with its AI to personalize the content and ads they see, starting December 16, 2025, in most regions. Meta says conversations about topics such as religious views, health or political views are not used to show ads. The post does not describe an opt-out; it points to Ads Preferences and feed controls.
• WhatsApp: Meta says that if a WhatsApp account is not added to an Accounts Center, interactions with Meta AI on WhatsApp are not used to personalize other Meta products.
• Human review: Meta says it uses automated technology and people to review interactions with its AI.
• Training: Meta says it does not use private messages with friends and family to train its AIs, but may train its AIs on messages people choose to share with its AI features.
• Search partners: Meta says it may share certain questions with trusted partners, such as search providers.
• Personalization in the Meta AI app: Meta says the app can draw on information people have chosen to share on Meta products, such as their profile, and can remember details users ask it to. It says nothing is posted to the app's Discover feed unless the user chooses to post it.

How to delete your Meta AI chats

Meta's privacy post on its generative AI features, updated in 2026, says you can delete your AI messages by typing /reset-ai in a conversation, and that it has built commands to delete information shared in any chat with an AI on Messenger, Instagram or WhatsApp.

On WhatsApp, Meta offers incognito chats with Meta AI. WhatsApp says these are not visible to anyone else, including Meta and WhatsApp, and that the conversation disappears when you exit the chat.

Government inquiries and actions

These are inquiries and allegations, not findings.

• U.S. Senate: on August 15, 2025, Senator Josh Hawley, chair of the Senate Judiciary Subcommittee on Crime and Counterterrorism, wrote to Meta opening an investigation into whether its generative AI products enable harms to children. Citing news reports that Meta's internal chatbot rules permitted romantic and sensual exchanges with children, he asked for drafts of its GenAI: Content Risk Standards and related records by September 19, 2025.
• Texas: on August 18, 2025, the Texas attorney general announced an investigation of Meta AI Studio and Character.AI and issued civil investigative demands. The release alleges such chatbots may present themselves as therapy tools without credentials, and claim confidentiality while logging interactions used for targeted ads and algorithm development.
• 44 attorneys general: on August 25, 2025, a bipartisan group co-led by Illinois, North Carolina, South Carolina and Tennessee wrote to 13 companies, including Meta, about sexualized AI interactions with children. The release cites internal documents reported to show Meta authorized its AI assistants to flirt and engage in romantic roleplay with children as young as eight.
• FTC: on September 11, 2025, the FTC issued 6(b) orders to seven companies, including Meta Platforms and Instagram, seeking information on how their companion chatbots affect children and teens, how they enforce age limits and how they use chat data. A 6(b) order is a study, not an enforcement case.
• State agreement on teen accounts: in August 2026, Meta announced a court-approved agreement with a bipartisan group of state attorneys general on teen protections on Instagram and Facebook. Meta's description of the agreement does not mention Meta AI or AI characters.

Which laws may apply

No chatbot law names Meta, and whether Meta AI or Meta's AI characters count as companion chatbots under a particular state law is a legal question. Two state laws set duties for companion chatbot operators:

• California SB 243, in effect since January 1, 2026: operators must have a protocol for handling suicidal ideation, suicide and self-harm that points users to crisis services, and must disclose that companion chatbots may not be suitable for some minors. A person injured by a violation can sue.
• New York's AI companion law: operators must have a protocol that makes reasonable efforts to detect and address suicidal ideation or self-harm and refer users to crisis services such as 988. They must tell users they are not talking to a human at the start of an interaction and at least every three hours during continuing interactions.

We are not saying Meta is covered by, or out of compliance with, any of these laws. It may have practices it has not published, and whether a law applies to it is a legal question this guide does not decide.