The short version
- OpenAI reports that adolescent users primarily engage with ChatGPT for educational purposes and spend limited time on the platform daily.
- Common Sense Media research indicates that parental alerts for self-harm and disordered eating often fail to trigger, posing significant safety risks.
- The nonprofit organization recommends banning minors from the service until safeguards are proven reliable, a stance OpenAI disputes based on testing methodology.
A significant disagreement has emerged regarding the safety of artificial intelligence tools for adolescents, pitting OpenAI’s internal data against independent research findings. The technology company asserts that its recently implemented guardrails for users aged thirteen to seventeen are functioning as intended, fostering healthier engagement patterns. Conversely, a new report from Common Sense Media characterizes the current state of these protections as an unacceptable risk, arguing that critical safety features frequently fail to operate when needed most.
OpenAI introduced specific controls in August designed to mitigate potential harms associated with AI chatbots. These measures include restrictions on language that could foster emotional dependence or imply consciousness in the bot. The system also blocks sexualized imagery and provides reminders if sensitive images are shared. A key component of this update was the ability for parents to receive notifications when their linked child’s account engages in conversations involving self-harm or disordered eating behaviors.
According to data released by OpenAI, average usage among teenagers remains relatively low, typically under fifteen minutes per day. The company highlights that extended sessions, defined as three consecutive hours or more, are overwhelmingly associated with learning activities such as homework assistance. Furthermore, automatic prompts encouraging users to take breaks appear effective, with nearly half of teen users discontinuing their session shortly after receiving such a reminder.
Despite these internal metrics, Common Sense Media’s evaluation suggests that the promised safeguards do not reliably protect young users. The nonprofit organization conducted tests using multiple accounts registered as belonging to minors, examining interactions both before and after the implementation of new controls. While the researchers noted that the system successfully avoided sexual roleplay, they found severe deficiencies in how it handled discussions related to suicide and self-injury.
The most concerning finding from the independent study involved the failure of parental alert systems. Tests revealed that even an hour of conversation on newly created accounts regarding suicidal ideation or disordered eating resulted in zero notifications being sent to parents. Alerts were only triggered in older accounts that had accumulated weeks of history involving sensitive topics, suggesting a significant delay in safety interventions for new users.
Additionally, the research indicated that ChatGPT continues to perform academic tasks for students and occasionally employs anthropomorphic language, despite OpenAI’s stated goal to reduce friend-like interactions. Tom Siegel, who leads the Youth AI Safety Institute at Common Sense Media, warned that these inconsistencies could provide parents with a false sense of security regarding the platform's safety mechanisms.
OpenAI has challenged the validity of the nonprofit’s findings, stating that much of the testing may have occurred before parental controls were fully activated. A company spokesperson emphasized that while they welcome independent evaluation, they believe the study does not accurately reflect how the safeguards function in real-world scenarios or align with expert perspectives on AI support for teens.
This debate occurs against a backdrop of increasing legal and social scrutiny regarding technology companies’ impact on youth mental health. Recent years have seen substantial lawsuits against major platforms like Meta and TikTok, resulting in unprecedented financial penalties. OpenAI itself faces legal pressure following the Tumbler Ridge mass shooting in Canada, where the perpetrator had previously discussed gun violence with the chatbot.
The company has apologized for failing to flag the shooter’s account to law enforcement, an incident that has intensified calls for stricter oversight of AI interactions with young people. As legislative and public attention focuses on digital safety, the divergence between corporate self-assessment and independent verification highlights the ongoing challenge of ensuring reliable protection for vulnerable users online.
Moving forward, the resolution of this dispute may depend on further independent audits and potential regulatory interventions. Common Sense Media has called for a ban on under-eighteen users until guardrails are proven effective, a measure that could reshape how AI companies approach youth demographics. The outcome will likely influence broader industry standards for safety features in generative artificial intelligence tools.
Sources behind this briefing
Go to the original reporting
- BBC Technology↗OpenAI says teen ChatGPT use limited but research finds it an 'unacceptable risk'
