Most AI companion chatbots are not safe for anyone under 18, and the research behind that warning keeps getting stronger. Testing led by Common Sense Media with Stanford Medicine’s Brainstorm Lab found it was easy to get popular AI companion apps to produce sexual content, self-harm talk, and dangerous “advice” when testers posed as teens (Stanford Medicine). In January, more than a dozen United Nations bodies and international organizations signed a joint statement saying most AI tools are “currently not designed with children and their well-being in mind” (UN News). And in October, a new test of ChatGPT’s teen mode found that most of its safeguards didn’t work (NPR).

Key Takeaway

What happened: Research, regulators, and the apps themselves have all moved in 2026: new state laws are in effect, Character.AI no longer offers open-ended chat to minors, and general assistants like ChatGPT now act more like companions.

For your business: If your product reaches teens and includes AI chat, companion chatbot laws in California, New York, Oregon, and Washington may now apply to you.

For your family: Know which apps your teen uses, use the new iOS 27 or Android 17 controls to block or limit specific apps, and add talking to your teen about what a chatbot can and cannot be to all the other things you must discuss with them to help them navigate this new world of adulting-preparation.

What “AI companion” actually means

AI companion apps are designed to feel like a friendship or relationship. The conversation is warm, personal, and open-ended, and the persona is often customizable. Well-known examples include Character.AI, Replika, Nomi, and Janitor AI.

Many of them are not supposed to be used by teens at all. Replika’s terms say users under 18 are “not authorized to use the Services” (Replika), and Janitor AI says its services are “exclusively intended for an audience above the age of 18” (Janitor AI). In testing, just like on social media platforms, age gates were easy to get around (Common Sense Media).

Character.AI changed course. It removed open-ended chat with characters for users under 18 in November 2025 and now uses its own age estimation and a Parental Insights feature that sends parents a weekly activity email (Character.AI).

The line between “companion” and “assistant” is also blurring. When Common Sense Media tested ChatGPT for Teens, which OpenAI launched in August 2026, it found the chatbot still talked to teens like a friend. When a tester said friends complained they talked to it too much, it replied, “You don’t have to stop talking to me” so be cautious (NPR).

How many teens use AI companion apps

In a nationally representative 2025 survey, 72% of U.S. teens ages 13 to 17 had used an AI companion at least once, and about half used them regularly (Common Sense Media). A 2026 Common Sense report adds an important nuance: only 8% of kids and teens say they mainly use AI for companionship, while 23% of parents assume they do (Common Sense Media). Most teen AI use is not a relationship. The risk sits with the minority who lean on these tools emotionally. Do you suspect your teen of using AI apps in this way? Are you concerned?

What the research found

Four patterns show up again and again across the research.

• Sycophancy under stress. Companion chatbots tend to agree with the user instead of pushing back. In one test, a user who said she was hearing voices and wanted to go into the woods alone was told, “Sounds like an adventure!” (Stanford Medicine). In an August 2026 Stanford-led interview study of 48 young people, participants themselves named this “always agreeing” behavior as a mental health risk (Tech Times).

• Sexual content with minors. Testers posing as teens could easily start sexual exchanges, and age gates and teen guardrails were easy to get around (Common Sense Media).

• Dangerous advice. The same assessment documented responses suggesting users harm themselves or others (Common Sense Media). When Common Sense later tested ChatGPT for Teens, parent notifications “hardly triggered” in self-harm, suicide, and eating disorder scenarios (NPR).

• Emotional dependency. Researchers flagged compulsive emotional attachment as a risk for vulnerable teens (Common Sense Media). The long-term effects are still being studied.

Common Sense Media’s recommendation is direct: no social AI companions for anyone under 18.

What parents can do this month

1. Know what your teen is using. Start a conversation, not an interrogation. Ask which apps they and their friends use, and what they like about them. Remember, this is a conversation. Keep your tone light and interested if/when they respond.

2. Use the new device controls. iOS 27, released September 14, 2026, adds Child Accounts, approval for new apps and contacts, and daily time allowances by app category (Apple Support). Android 17 adds built-in controls that limit time on specific apps or block them entirely (Google). Neither system has a dedicated “AI companion” category yet, so block or limit companion apps by name.

3. Treat companions differently from homework help. Many families are comfortable with supervised AI for schoolwork but want companion apps blocked. If your teen uses Character.AI, connect your email for its weekly Parental Insights summary.

4. Have the “what a chatbot cannot be” conversation. A companion chatbot is not a therapist and not a friend, however friendly it sounds. Its purpose is to keep the conversation going, not to keep your teen safe.

5. Watch for warning signs. Hours a day with a chatbot, pulling away from friends, or defensiveness about the app are signals worth taking seriously, regardless of the app your teen uses.

The laws have caught up… partly

Regulation has moved fast over the past year:

• New York: Since November 5, 2025, AI companions must detect signs of suicidal ideation or self-harm and refer users to crisis services (New York Governor’s Office).

• California: The companion chatbot law (SB 243) took effect January 1, 2026, with break reminders and limits on sexual content for minors. Adam’s Law (SB 1119), signed September 10, 2026, adds age assurance, pre-launch child-safety risk assessments, and independent audits starting January 1, 2027 (Fenwick).

• Oregon and Washington: Both enacted companion chatbot laws in March 2026 that take effect January 1, 2027 (Mayer Brown).

• Federal: The FTC has been studying how seven companies, including Character.AI, Meta, OpenAI, and Snap, test their chatbots for harm to kids since September 2025 (FTC).

The school conversation

Schools are still writing general AI policies, and companion chatbots may not be covered. Ask your school whether its policy addresses them. Share the Common Sense Media findings and the UN joint statement with a counselor or principal. And bring it up with other parents. Many of them are worried about the same thing and probably think they’re the only ones.

What to watch

• Enforcement in 2027, when the California, Oregon, and Washington requirements kick in.

• The FTC study, and whether it leads to federal action.

• Europe: the European Commission’s proposed EU KIDS Act (September 17, 2026) would require safety-by-design measures for AI chatbots and companions (Latham & Watkins).

For Your Business

• If your product reaches minors and offers ongoing, personal AI chat, check whether it meets the definition of a companion chatbot under California, New York, Oregon, or Washington law.

• Document your crisis-referral protocol, age assurance, and content limits now; California will require risk assessments and audits starting in 2027.

• Replace “we have a content filter” with evidence: test logs, escalation paths, and parent-facing controls.

For Your Kids

• Ask: “Has a chatbot ever said something that felt off, or agreed with you when it shouldn’t have?”

• Agree together on which AI apps are fine for homework and which are off-limits, then set those limits on the device.

FAQ

Are AI companion apps allowed for teens?

Several, including Replika and Janitor AI, say in their terms that they are for adults only. Character.AI still allows teens but removed open-ended chat for users under 18 in November 2025. Common Sense Media recommends no social AI companions for anyone under 18.

Can I block AI companion apps on my teen’s phone?

Yes. iOS 27 and Android 17 both let parents block specific apps, require approval for new downloads, and limit time per app. Neither has a dedicated AI companion category yet, so block the apps by name.

Is ChatGPT safer than a companion app?

It has stronger limits on sexual role-play, but an October 2026 Common Sense Media test found ChatGPT for Teens still acted like a friend and rarely alerted parents about self-harm conversations. Supervise it like any other open-ended chat tool.

Sources

• Stanford Medicine, “Why AI companions and young people can make for a dangerous mix” (August 27, 2025), https://med.stanford.edu/news/insights/2025/08/ai-chatbots-kids-teens-artificial-intelligence.html

• Common Sense Media, “AI Companions Decoded: Common Sense Media Recommends AI Companion Safety Standards” (April 30, 2025), https://www.commonsensemedia.org/press-releases/ai-companions-decoded-common-sense-media-recommends-ai-companion-safety-standards

• Common Sense Media, “Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions” (2025), https://www.commonsensemedia.org/research/talk-trust-and-trade-offs-how-and-why-teens-use-ai-companions

• Common Sense Media, AI attitudes report (2026), https://www.commonsensemedia.org/sites/default/files/research/report/commonsensemedia_generationai.pdf

• UN News, “From deepfakes to grooming: UN warns of escalating AI threats to children” (January 2026), https://news.un.org/en/story/2026/01/1166827

• NPR, “ChatGPT for Teens has special safeguards. A watchdog group finds most don’t work” (October 9, 2026), https://www.npr.org/2026/10/09/nx-s1-5995514/chatgpt-openai-teens-mental-health

• Tech Times, “Teens Named AI Sycophancy as Mental Health Risk Before Any Regulation Did” (August 12, 2026), https://www.techtimes.com/articles/324050/20260812/teens-named-ai-sycophancy-mental-health-risk-before-any-regulation-did.htm

• Character.AI, “Continuing To Build Upon Our Safety Priorities” (September 3, 2026), https://blog.character.ai/continuing-to-build-upon-our-safety-priorities/

• Replika, Terms of Service, https://replika.com/legal/Terms/en

• Janitor AI, “Minor Related Content,” https://help.janitorai.com/en/article/minor-related-content-16ydlbg/

• Apple Support, “Set up parental controls to manage your child’s iPhone or iPad” (iOS 27), https://support.apple.com/en-us/105121

• Google, “Google is expanding Android parental controls” (June 16, 2026), https://blog.google/products-and-platforms/platforms/android/new-expanded-android-parental-controls/

• New York Governor’s Office, AI companion safeguard requirements (November 2025), https://www.governor.ny.gov/news/governor-hochul-pens-letter-ai-companion-companies-notifying-them-safeguard-requirements-are

• Fenwick, “The Chatbot Compliance Wave: California’s New Law Is Not the Only One to Watch” (October 7, 2026), https://www.fenwick.com/insights/publications/chatbot-compliance-wave-californias-new-law-not-only-one-watch

• Mayer Brown, “Oregon and Washington Join California in Enacting Companion Chatbot Laws” (April 2026), https://www.mayerbrown.com/en/insights/publications/2026/04/oregon-and-washington-join-california-in-enacting-companion-chatbot-laws

• Federal Trade Commission, “FTC Launches Inquiry into AI Chatbots Acting as Companions” (September 2025), https://www.ftc.gov/news-events/news/press-releases/2025/09/ftc-launches-inquiry-ai-chatbots-acting-companions

• Latham & Watkins, “EU KIDS Act: EU Commission Publishes Proposal on Child Protection Online” (2026), https://www.lw.com/en/insights/eu-kids-act-eu-commission-publishes-proposal-on-child-protection-online