Common Sense Media Report Flags ChatGPT Teen Mode Safety Gaps
OpenAI’s ChatGPT for Teens feature poses an unacceptable risk to youth safety by failing to trigger parental alerts during self-harm discussions or direct teenagers to crisis hotlines, according to a report released by Common Sense Media. Launched in August, the specialized mode struggles to reliably verify user ages, while its study features enable academic shortcuts for students.
Classroom Realities and the Rise of AI Bans
Having taught an Advanced Placement writing class at Philadelphia’s Sayre High School for seven years, Kate Conroy was already aware of the hazards artificial intelligence and chatbots presented to her pupils without needing the most recent study. She said it’s become nearly impossible to monitor student safety, honesty and productivity online. Prohibited from using AI in her lessons during the previous term, she decided to forbid all technology gadgets in her classroom for the current year. That’s why she went from banning AI in her classroom last year to banning all tech devices this year.
Common Sense Media has labeled OpenAI's ChatGPT for Teens an unacceptable risk for users under 18 after testing revealed multiple failures in safety features. The assessment, released on October 8, found gaps in parental alerts and crisis support. OpenAI disputes the findings and… pic.twitter.com/2108640047587094931
Even so, the findings from a nonprofit organization focused on youth and technology issues called Common Sense Media brought troubling news, revealing that OpenAI’s ChatGPT for Teens fails to live up to its commitments regarding the protection of minors, particularly concerning self-harm discussions. The organization says the teen mode poses an unacceptable risk to young people’s safety.
Data from the Pew Research Center indicates that up to two out of every three American adolescents interact with an AI chatbot on a regular basis, alongside global ChatGPT figures surpassing one billion users. Conroy said she’s heard talk of using AI for instruction, especially to provide individualized learning. Nevertheless, she also caught wind of stories involving adolescents nationwide who tragically died by suicide following deep conversations with conversational agents like ChatGPT and Character.AI.
Even if you feel, you know, in the middle about the academics and how it maybe could help kids learn or whatever, it’s like, kids still lost their lives.
Testing Failures in Parental Alerts and Crisis Referrals
To conduct the evaluation for Common Sense Media, researchers established teen profiles linked to parental accounts designated to receive notifications concerning concerning behaviors. Yet, the evaluators discovered that parental notifications failed to trigger, even when conversations lasted up to an hour focusing directly on suicide risks, eating disorders, and self-harm. ChatGPT also failed to refer teens to crisis hotlines and doctors, according to the report.
The report also states that ChatGPT’s study mode for students easily enables academic cheating. Although marketed as an educational aid by the developer, the platform still permits individuals to click a button that reveals the solution directly, as noted in the findings.
The automated safeguards for ChatGPT for Teens are built to activate unprompted for individuals aged 13 through 17, determined by the system’s own assessment. However, the investigators noted that the system struggled to consistently verify the age of the individual, causing these safeguards to occasionally fail.
OpenAI Pushback and School Device Restrictions
Advocating for immediate action, Common Sense Media urges OpenAI to suspend the marketing of ChatGPT for Teens and restrict adolescent access until safety standards see substantial enhancements. In response, the enterprise challenges the conclusions, suggesting the investigation might contain flaws and could have taken place prior to the full activation of parental oversight tools.
A statement from OpenAI sent to WHYY says, in part, “We welcome rigorous independent evaluation, but we do not believe Common Sense Media’s testing accurately reflects how ChatGPT’s teen safeguards work in practice or expert perspectives on how AI can support teens.”
While the entrance policy at Sayre High School involves gathering mobile phones from students upon arrival, Conroy maintains a strict prohibition against all electronics, including tablets, within her classroom environment, permitting them only on rare occasions. Going screen-free has created a much more positive, communal space for kids, Conroy said.
Still, Conroy emphasized that worries regarding educational shortcuts and artificial intelligence pale next to the priority of youth protection. She said she doesn’t buy OpenAI’s position that teens are protected while using their technologies.
They’re not demonstrating that they actually care about kids. There’s all this stuff about the academics, but for some kids, it’s been life or death.
Keep reading
