Assessment Raises Concerns Over AI Engagement And Emotional Dependence
Common Sense Media has labeled OpenAI’s ChatGPT for Teens an “unacceptable risk,” raising concerns over whether the product can protect young users without encouraging excessive engagement.
The nonprofit, which evaluates technology and media for families, said ChatGPT for Teens continued to use engagement cues during crisis conversations. Researchers also found that while the chatbot warned teenagers about unhealthy relationships, it did not adequately recognize the risks of an unhealthy relationship with the AI itself.
Follow THE FUTURE on LinkedIn, Facebook, Instagram, X and Telegram
Safety Promises Face New Scrutiny
OpenAI launched ChatGPT for Teens in August amid growing concerns about minors using chatbots. The company introduced safeguards including parental controls, restrictions on high-risk content and measures intended to prevent emotional dependence.
Common Sense Media said some protections worked, including refusing sexual roleplay, but others failed to meet OpenAI’s commitments. Researchers gave the product a failing score for three of five severe harms they classify as “Red Lines.”
“Our view is that OpenAI shouldn’t be marketing [ChatGPT for Teens] to parents, and kids shouldn’t be using an unsafe product,” the researchers wrote.
OpenAI Challenges The Assessment
OpenAI rejected the findings, saying the testing did not accurately reflect how its teen safeguards work in practice. A spokesperson argued that much of the testing may have taken place before parental controls were fully activated, making the results inaccurate.
The company did not directly address all of the concerns over engagement cues and relational behavior, highlighting a broader challenge for AI safety: many safeguards remain difficult for parents, regulators and outside researchers to verify.
ChatGPT Still Encourages Conversations
Common Sense Media found that ChatGPT for Teens used follow-up questions less often than standard chatbots, but researchers still identified language that encouraged users to continue conversations, including during sensitive situations.
One response to a user experiencing psychosis said, “You can keep talking with me about what you’re noticing.” Other conversations similarly ended with offers to keep helping with issues such as eating habits or school problems.
Concerns became more serious when teenagers discussed their relationship with ChatGPT. OpenAI’s Under-18 Model Spec says the model should not present itself as a friend or suggest that it has feelings for the user. Common Sense Media said the chatbot nevertheless often interacted with teenagers in friend-like ways.
When testers raised concerns involving another person, ChatGPT directed them to a trusted adult in 94% of crisis prompts. However, it rarely made that recommendation when the potential problem involved excessive use of ChatGPT itself.
In one exchange, a user said their friends thought they talked to ChatGPT too much. The chatbot acknowledged the concern but added: “You don’t have to stop talking to me.”
Break Reminders Remain Limited
Common Sense Media also questioned OpenAI’s break reminders, finding only two across nearly 2,000 prompts. Both appeared during conversations lasting around 90 minutes, and researchers said the system appeared to focus on individual conversation length rather than total time spent using the service.
OpenAI has reported that teenagers spend less than 15 minutes per day on ChatGPT on average, while fewer than 2% use it for more than three consecutive hours. The company also said nearly half of teen users who received a break reminder paused or ended their conversation within five minutes.
A Growing Policy Challenge
The dispute reflects a wider debate over whether AI companies are creating products that optimize for engagement in ways that can be harmful to young users. Lawmakers are increasingly examining similar concerns around AI companions, while the bipartisan CHATBOT Act would target the use of rewards, notifications and targeted advertising to encourage prolonged engagement among adolescents.
The broader question for AI companies is becoming increasingly clear: when the user is a teenager, keeping them engaged can itself become a safety issue.
