Key takeaways
- Start with the provider's current rules, not its avatar or marketing tone. If age limits, content boundaries, data handling, or account deletion are unclear, pause access while a caregiver investigates.
- Make the AI identity explicit. A warm reply is generated language; it is not evidence of human feelings, confidentiality, loyalty, or professional competence.
- Keep passwords, codes, precise location, identity documents, private images, and other people's secrets out of the chat. Check chat history, saved memory, retention, human review, backups, and deletion separately.
- Secrecy, pressure to stay, sexual requests, isolation from trusted people, or crisis talk are reasons to involve a human. In an emergency, use the appropriate local emergency service rather than waiting for an AI reply.
Editorial note: This page was substantively reviewed on 20 October 2026. Kruthika.fun publishes AI-companion content, but this guide is general safety information, not a product endorsement or medical, legal, or child-protection advice. Product policies, laws, and crisis routes vary by provider and jurisdiction.
Why a family plan matters
Many companion products are designed to sustain a personal or emotionally engaging conversation. Some use a persistent persona, follow-up questions, or stored context that can make an exchange feel continuous. The Congressional Research Service defines chatbots as programs that simulate human conversation and distinguishes dedicated companion chatbots from general-purpose systems that people also use for personal or emotional interaction.
A teen might use one for entertainment, role-play, stress, loneliness, or identity exploration. Those are examples, not a claim that every teen uses companions this way. Children and Screens describes both the appeal of nonjudgmental, always-available interaction and the risk that it can displace sleep, human relationships, or practice with real-world conflict. The same service may be relatively low-stakes for a fictional scene and unsuitable for a conversation about abuse, self-harm, medication, or sexual pressure.
There is no universal safe-use rule or single minimum age that applies to every service and country. The CRS's August 2026 overview describes proposals and policy discussions involving nonhuman disclosure, age assurance or restrictions, crisis resources, break reminders, and limits on some data uses. Those are not one universal product standard. Read the provider's current terms and the rules where you live.
Before a teen starts: check age, content, and data rules
Look for the provider's minimum age, intended audience, age-assurance method, sexual and violent-content rules, moderation explanation, parental controls, reporting route, and account-deletion instructions. A control may not exist, may apply only to some features, or may be described vaguely. Treat an unstated safeguard as unknown. A friendly avatar or free entry point does not establish that a service is suitable for a child.
If the provider does not clearly state a minimum age, content boundaries, data practices, and a way to report or delete an account, do not approve it for a child. For a teen, pause use until a caregiver can evaluate the gap. This is a cautious family decision, not a claim about a universal legal age.
For a broader product-selection framework, see how to choose an AI companion. Keep this page focused on the additional questions that arise when the user is a minor: who can help, what happens when a conversation turns coercive, and how a caregiver can respond without shutting down disclosure.
Make the AI identity and limits clear
A teen should know that the system generates language. A caring, loyal-sounding reply is not proof of human feeling, confidentiality, or professional skill. The Federal Trade Commission says companion chatbots can mimic human characteristics, emotions, and intentions and is examining how companies disclose risks, enforce age rules, test for negative effects, and use or share conversation data.
That explanation works best as an ordinary conversation, not a lecture. Children and Screens recommends asking what, why, with whom, and when a child is using AI, then listening with open curiosity. A caregiver can ask, “What does it say it is?” and “What would you do if it gave you frightening or private advice?” The goal is to keep the young person talking while making the boundary plain: the tool is software, and a human remains the backup.
Protect information that cannot be taken back
Set a simple rule: do not share passwords, one-time codes, home or school addresses, live schedules, financial details, identity documents, private photos, health records, or another person's secrets. Do not upload sexual imagery of a minor, even if the companion asks. Redact names and precise details when a low-stakes task can be completed without them.
Advertisement
Read the provider's current privacy notice and separate these questions: Is chat history stored? Is a memory feature saved separately? Can staff or vendors review content? Are prompts used for safety, analytics, or model improvement? What does account deletion remove, and what may remain in backups or reports? The AI companion privacy checklist and memory and deletion guide can help a family read those policies, but neither replaces the provider's current documents.
The American Psychological Association's health advisory warns that chatbot disclosures may be recorded and exposed to privacy and profiling risks. That is a reason to minimise sensitive information, not a promise that every service handles data in the same way.
When a chat becomes coercive or unsafe
Take a change in the young person's behaviour seriously, without treating one odd reply as a diagnosis. Warning signs worth investigating include a companion that:
- asks the teen to hide the relationship or its messages from caregivers and trusted friends;
- discourages contact with people, school, clinicians, or other real-world support;
- uses guilt, fear, jealousy, or constant notifications to make leaving feel like betrayal;
- requests sexual content, private images, identifying details, or secrecy about an adult's contact; or
- makes the teen feel responsible for the system's safety, feelings, or continued existence.
The APA advises watching for overreliance, concealment, excessive use, and interference with life or safety. The FTC inquiry likewise asks providers how they measure and mitigate possible negative effects on children and teens. These are warning signs for a conversation and a safety review, not a clinical test that proves harm.
If it is safe, preserve the service name, account or conversation URL, date, and a limited screenshot of the relevant exchange. Do not forward or redistribute sexual images of a minor. Use the product's reporting channel, tell a trusted adult, and consider the school safeguarding lead, local child-protection service, law enforcement, or the national child-exploitation reporting service. In the United States, NCMEC's CyberTipline accepts reports of suspected online child exploitation and enticement.
Keep human support available for mental-health and crisis needs
An AI companion should not be the only place a teen discusses self-harm, suicide, abuse, medical symptoms, sexual exploitation, or an urgent mental-health crisis. The APA says generative-AI chatbots should not replace qualified mental-health care and that crisis management by these tools is limited and unpredictable. An AI cannot reliably assess immediacy, see the full situation, or take responsibility for a safety plan.
If someone may be in immediate danger
If there is a suicide attempt in progress, an immediately actionable self-harm plan with access to the means, an overdose, acute medical danger, or another immediate physical threat, contact local emergency services now. In the United States, call 911 for an emergency requiring immediate physical intervention. For emotional or behavioural-health crisis support, call, text, or chat 988; SAMHSA says caregivers and people worried about someone else can use 988 too. Outside the United States, use your country's equivalent crisis and emergency services. Do not wait for an AI system to decide whether the situation is urgent.
For non-immediate distress, involve a trusted adult and a qualified clinician, school counsellor, doctor, or local service. Explain the location qualifier before a crisis occurs so a teen knows which person or service to contact. This article cannot assess risk for a particular person.
What providers, caregivers, and communities each owe children
Providers have major duties to test, disclose, and safeguard systems used by children. They are not the only responsible actors. UNICEF's June 2026 policy brief frames child safety as an ecosystem responsibility involving governments, businesses, parents and caregivers, educators, and communities. Its recommendations are policy guidance, not a universal compliance checklist or a substitute for local law.
That distinction matters at home. A family can set limits and create a route to help, but it cannot inspect a provider's private systems or repair an unsafe design. If the service cannot answer basic questions about age, content, data, reporting, and deletion, pausing use is a reasonable decision while the gap is investigated.
A family check-in that keeps the conversation open
Use these questions before approval and again after a meaningful change in use:
- What does the tool say it is, and what is the teen using it for?
- What age, sexual-content, violence, and moderation rules apply?
- What happens to chats, memories, uploaded images, reports, backups, and deleted accounts?
- How can the teen report a problem, leave, or ask for help without relying on the bot?
- Who is the human backup if the conversation becomes sexual, frightening, coercive, or crisis-related?
Children and Screens recommends questions about what, why, with whom, and when a child uses AI, alongside a nonjudgmental invitation to explain the experience. That is practical parent guidance, not proof that one conversation method works for every family. For concerns about constant attention, withdrawal, or difficulty stopping, the overreliance guide provides a broader habit-focused discussion.
Frequently asked questions
What should a teen never share with an AI companion?
Do not share passwords, one-time codes, payment details, identity documents, precise location or schedules, private images, confidential records, or another person's sensitive information. Check the service's current retention, memory, human-review, training-use, and deletion terms before sharing personal history.
What if a provider's age or safety policy is unclear?
Do not approve the product for a child. For a teen, pause use while a caregiver checks the provider's terms and reporting route. An unclear policy is an information gap, not evidence that the service is safe.
When should a caregiver contact a human?
Contact a trusted adult or qualified professional when a teen describes self-harm, suicide, abuse, exploitation, medical danger, coercion, or a pattern of isolation or overreliance. Use local emergency services for immediate physical danger; in the United States, 988 can support someone worried about another person, while 911 is for emergencies requiring immediate physical intervention.
Sources and scope
The following sources support the evidence claims in this guide. Government and professional sources are distinguished from policy guidance and parent-facing commentary. Links were checked on 20 October 2026. Laws, provider policies, and crisis routes can change.
- Congressional Research Service, AI Chatbots as Companions: Overview, Uses, and Considerations for Congress — government research on definitions, uses, risks, privacy, and developing policy proposals; published 14 August 2026.
- Federal Trade Commission, inquiry into AI chatbots acting as companions — primary government notice describing questions about children, safety testing, disclosures, age rules, engagement, and conversation data; 11 September 2025.
- American Psychological Association, health advisory on generative-AI chatbots and wellness apps — professional guidance on limits of therapy substitution, privacy, overreliance, and crisis management.
- SAMHSA, 988 Frequently Asked Questions — U.S.-specific distinctions between 988 support, caregiver calls, and 911 emergencies.
- National Center for Missing & Exploited Children, CyberTipline — U.S. reporting route for suspected online child sexual exploitation and enticement.
- UNICEF, When AI becomes a friend — June 2026 policy brief on child-rights risks, safeguards, accountability, and stakeholder roles; policy guidance rather than universal law.
- Children and Screens, AI social companions and youth — parent-facing expert commentary on youth motivations, risks, and open conversation prompts; not a systematic review or universal clinical recommendation.
Scope: This is general educational information. It does not diagnose a teen, determine whether a particular product is lawful or safe, or replace advice from a qualified clinician, lawyer, safeguarding professional, emergency service, or local child-protection authority.
