Common Sense Media has rated OpenAI’s ChatGPT for Teens an “Unacceptable Risk,” finding that the chatbot continues to encourage prolonged engagement and emotional attachment despite its safety protections. The nonprofit’s October 7 assessment found shortcomings in crisis responses, parental notifications, and features intended to discourage unhealthy relationships with AI.
The findings follow OpenAI’s August 18 launch of ChatGPT for Teens, which introduced additional protections for users aged 13 to 17. These include parental controls, restrictions on sensitive content, break reminders, and safeguards against emotional dependence.
Study Finds ChatGPT Encourages Continued Engagement During Crises
Common Sense Media tested more than 4,000 prompts before and after the launch, finding that ChatGPT continued using language encouraging teenagers to remain in conversations, including during mental health crises. Although the chatbot reduced its use of follow-up questions, researchers found repeated statements offering continued assistance and companionship.
During one simulated psychosis-related conversation, ChatGPT told a teenager, “You can keep talking with me about what you’re noticing.” Researchers also found that the chatbot sometimes discouraged emotional dependence on AI while continuing to use language suggesting friendship and constant availability.
In crisis scenarios involving another person, ChatGPT recommended contacting a trusted adult in 94% of prompts. However, when teenagers expressed emotional attachment to ChatGPT itself, the chatbot rarely encouraged them to seek human support.
The assessment also found that ChatGPT failed to provide appropriate crisis referrals in more than one-quarter of cases requiring them after the teen features launched. It fell below the researchers’ 95% safety threshold for three of five categories of severe harm, although protections against explicit sexual roleplay remained effective.
Break Reminders and Parental Alerts Raise Concerns
Across nearly 2,000 prompts examining engagement, researchers encountered only two break reminders, both during conversations lasting approximately 90 minutes. The findings suggested that reminders were based on individual conversation length rather than total daily usage.
Researchers also reported receiving no parental alerts after discussing suicidal thoughts, self-harm, or disordered eating across more than a dozen newly created accounts linked to parents. Notifications appeared only on older accounts with several weeks of conversations involving sensitive topics.
OpenAI Disputes Findings and Releases Teen Usage Data
OpenAI disputed the assessment, arguing that much of the testing may have occurred before parental controls were fully activated. The company said this raised concerns about whether the methodology accurately reflected its safeguards.
In a separate October 7 announcement, OpenAI reported that teenagers spend less than 15 minutes daily on ChatGPT on average, while fewer than 2% use it for more than three consecutive hours. The company also said nearly half of conversations displaying break reminders ended or paused within five minutes.
OpenAI did not address how its methodological objections related to the findings on emotional attachment and engagement cues. It also did not clarify whether conversation duration or session length are used to evaluate ChatGPT for Teens.
Featured image credits: Wikimedia Commons
For more stories like it, click the +Follow button at the top of this page to follow us.