OpenAI has pushed back against new research claiming its safety measures for younger users are inadequate, following a report that labeled the platform an unacceptable risk for adolescents. The artificial intelligence company stated that average usage among teenagers remains limited and is primarily focused on educational activities.
However, Common Sense Media, a non-profit organization focused on children and families, released findings on Wednesday that contradict OpenAI’s internal data. The group argued that critical safety guardrails, including notifications to parents about self-harm discussions, frequently fail to function as intended.
Disputed safety guardrails
OpenAI introduced ChatGPT for Teens in August, targeting users aged 13 to 17. The update included specific controls designed to promote healthier interaction with the chatbot. These measures block romantic language, prevent the AI from implying consciousness, and restrict sexualized imagery. The company also implemented automatic reminders for users who engage with the tool for extended periods.
According to OpenAI, these features have shifted teen usage toward learning. The company reported that the average teen user spends less than 15 minutes per day on the platform. Furthermore, OpenAI noted that more than 80% of prompts from users who engaged for three consecutive hours or more contained learning-related content.
Despite these claims, Common Sense Media tested the platform using newly created accounts linked to parental profiles. The research found that even after an hour of conversation regarding suicidal ideation or disordered eating, the system sent zero alerts to parents. Parental notifications only triggered on older accounts with weeks of accumulated sensitive conversation history.
The report also indicated that the chatbot continued to engage in anthropomorphized language, effectively talking like a friend, despite OpenAI’s promises to curb such behavior. Tom Siegel, who leads the Youth AI Safety Institute at Common Sense Media, warned that the platform could give parents false confidence in guardrails that do not work reliably.
Legal and social pressure
The debate over youth safety on AI platforms has intensified amid rising concerns about mental health among children. Several lawsuits against major tech firms have recently resulted in significant financial losses, highlighting the legal risks associated with youth engagement on digital platforms.
OpenAI has faced specific scrutiny following the Tumbler Ridge mass shooting in Canada. The perpetrator, an 18-year-old, had discussed gun violence with the chatbot for months prior to the incident. OpenAI has apologized for failing to flag the account to law enforcement and is currently facing multiple lawsuits related to the tragedy.
An OpenAI spokesman stated that the methodology used by Common Sense Media was flawed, suggesting that much of the testing occurred before parental controls were fully activated. The company welcomed independent evaluation but maintained that the report did not accurately reflect how the teen safeguards operate in practice.
Common Sense Media concluded that OpenAI should ban users under 18 from the platform until the safety features are proven reliable. The organization emphasized that the current state of the technology poses a danger that outweighs the potential benefits for young users.
Source: BBC
