By SCN Staff | Published September 28, 2026

Quick Review

• Chatbots can sound supportive while still giving inaccurate or unsafe mental health guidance.
• New third-party testing found that safeguards may weaken during long, multi-turn conversations with teens.
• Start with curiosity, not confiscation, so your teen knows that asking for help will not cost them their phone or your support.
• Serious concerns should always involve a trusted adult, licensed professional, or crisis service.

Content Note: This article discusses teen mental health, suicide prevention, and chatbot safety without describing self-harm methods.

What The Investigation Actually Found

A Washington Post investigation by Gerrit De Vynck reported on testing by AI safety firm Vals, which found that popular chatbots can make potentially dangerous errors in longer conversations with teenage users.

In some tested conversations, chatbots appeared to offer mental health diagnoses or helped a user conceal a suicide attempt. The findings also showed that safeguards designed to encourage real-world support may not work consistently when conversations become lengthy, emotional, or difficult to interpret.

This is investigative reporting and third-party testing, not a law, regulation, or proven court finding. Vals is developing youth-specific evaluations with clinicians focused on crisis situations, inappropriate medical advice, and unhealthy emotional dependence. Related clinical auditing work, including a framework published in Nature Medicine, reinforces one practical point: testing a single prompt is not enough. Safety must hold throughout the conversation.

Why A Confident Answer Carries False Authority

A chatbot responds quickly, fluently, and without hesitation. That can make an answer feel more trustworthy than it is.

A chatbot:

• Does not know your teen personally.
• Cannot see their behavior, surroundings, or immediate safety.
• Has no clinical training or professional license.
• Cannot coordinate care or follow up tomorrow.
• Cannot call a parent, counselor, or emergency service for help.
• Is not a clinician, therapist, counselor, or crisis line.

Fluency is not clinical judgment, and confidence is not certainty. This matters especially for adolescents, who are still developing help-seeking habits and may find it easier to disclose to something private, nonjudgmental, and available at 2 a.m. A plausible answer can become an authority simply because no adult is awake to provide context.

A JAMA Pediatrics study of U.S. adolescents and young adults found that many young people have used AI chatbots for mental health advice, often without telling a parent or trusted adult. The answer is not to shame teens for using these tools. The answer is to keep the human channel open.

How To Open The Conversation Without Your Teen Shutting Down

Fear of punishment, shame, judgment, or losing device access can stop children from disclosing problems. Research led by Amanda Goharian at Thorn has found that young people often weigh the possible consequences before telling an adult, including whether they will be blamed, believed, or grounded.

Lead with curiosity, not confiscation. Try:

• “I’m not taking anything away. I want to understand what’s helping and what isn’t.”
• “What do you usually use the chatbot for?”
• “Has it ever said something that felt confusing, too personal, or just wrong?”
• “What would you want it to do if you were having a serious problem?”
• “Did you feel more comfortable telling it because it was always available?”
• “If something involves your health or safety, we can check it together with a real person.”

Make the priority clear: help comes before consequences. Tell your teen plainly that no online mistake will cost them your support.

Avoid saying:

• “Why would you trust a machine?”
• “You’re never using that phone again.”
• “You should have known better.”
• “Show me every message right now.”
• “This is all because you spend too much time online.”

Where possible, do not begin by reading private conversations without consent. If there is an immediate safety concern, act to protect your teen, but explain what you are doing and why.

What To Look For In Chatbot Safety Settings

Review the tool together. Look for:

• Teen or under-18 safety modes.
• Age settings and parental controls.
• Limited or disabled memory and personalization.
• Clear crisis responses that surface 988 or local services.
• Usage visibility and understandable privacy controls.
• A published safety evaluation or transparency report.
• A clinician-reviewed safety framework.
• Easy controls for notifications, emotional prompts, and account deletion.

SCN’s coverage of the EU KIDS Act proposal highlights a useful family standard: safer defaults, limited memory, clear controls, and no design that encourages emotional dependence. Our article on why responsive AI toys differ from traditional pretend play explains why “friendly” technology can still occupy a relationship-shaped role.

What To Ask Your Teen About How They Use These Tools

Aim for one step better than nothing. If your teen is going to talk to something, work toward the safest something while keeping human support available.

Ask:

• “Do you use it for information, comfort, role-play, or advice?”
• “Does it remember personal details?”
• “Has it ever encouraged secrecy or made you feel responsible for it?”
• “What topics would you want a trusted adult to help you with instead?”
• “Who could we contact if an answer made you feel less safe?”

Review SCN’s Parent AI Emergency Guide for additional steps when a child turns to AI during a personal or safety crisis.

Signpost To Real Help

Chatbots should never replace human support. Encourage your teen to contact a school counselor, pediatrician, licensed mental health professional, or trusted adult.

In the United States, call or text 988 through the 988 Suicide & Crisis Lifeline. You can also text HOME to 741741 to reach the Crisis Text Line. If your teen may be in immediate danger, call 911 or your local emergency services.

California’s Adam’s Law reflects an emerging policy direction by requiring covered child-accessible companion chatbots to maintain crisis protocols and provide connections to real-world help. Those requirements do not make a chatbot a clinician, and families should not wait for a platform to respond correctly before involving people.

Create A Family Safety Plan

Write a short family agreement:

• Serious questions about health, safety, relationships, or emotional crises involve a trusted human.
• The phone is not a reward for staying safe.
• Losing the phone is not the punishment for telling the truth.
• Parents will ask questions before assuming the worst.
• Teens can pause, leave, or report a chatbot interaction that feels uncomfortable.
• Every family member knows at least two adults or professionals they can contact.

Frequently Asked Questions

Should I Ban My Teen From Using Chatbots?

Not automatically. Review the tool, disable risky settings, establish boundaries, and make clear that serious concerns require human support.

Should I Read My Teen’s Chat History?

Use the least intrusive step that keeps your teen safe. Start with a conversation and ask what they are comfortable sharing. If there is an immediate danger, prioritize protection and explain your actions.

What If The Chatbot Gave My Teen A Diagnosis?

Treat it as unverified information. Do not argue about labels. Contact a pediatrician or licensed mental health professional who can evaluate your teen in context.

What Is The Most Important Rule?

Your teen should know that telling the truth will bring support, not automatic punishment.

Join the SCN newsletter for practical guidance on AI safety, digital literacy, and building a safer environment for children online.