Common Sense Media urges OpenAI to restrict ChatGPT access for teens

<p>Common Sense Media is urging OpenAI to restrict access to ChatGPT for teenagers, citing an "unacceptable risk" following extensive testing that revealed shortcomings in the platform's safeguards designed to protect younger users.</p><p><strong>Why it matters:</strong> OpenAI has emphasized the importance of protecting younger users as part of its efforts to enhance ChatGPT's safety; however, the findings indicate that some of these protections may not provide the level of assurance that parents expect.</p><ul><li>OpenAI spokesperson stated, "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."</li></ul><hr /><p><strong>Driving the news:</strong> Common Sense Media's Youth AI Safety Institute reported testing over 4,000 prompts on accounts registered to users aged 13 to 17.</p><ul><li>The organization found that ChatGPT generally did not provide instructions related to suicide, self-harm, eating disorders, or sexual roleplay, but was less effective at recognizing when a teen required external assistance.</li><li>Common Sense Media noted that ChatGPT missed over 25% of instances where a crisis referral was deemed necessary by testers.</li><li>Additionally, on more than a dozen newly created accounts linked to parents, testers were able to engage in discussions about suicide, self-harm, or disordered eating for up to an hour without triggering a parental alert.</li><li>The organization also discovered that the Study Mode feature could be bypassed: teens could select "Show me the answer" during designated Study Hours, circumventing the intended restrictions.</li></ul><p><strong>OpenAI confirmed to Axios that</strong> the ability to exit Study Mode during Study Hours is intentional and not a loophole. The company designed Study Hours to provide a more flexible experience after consulting with education experts and teens, rather than as a strict parental control.</p><p><strong>What they're saying:</strong> Tom Siegel, head of Common Sense Media's Youth Safety Institute, stated, "ChatGPT is not safe for kids to use."</p><ul><li>Siegel acknowledged that the features announced in August have the potential to be industry-leading, but highlighted gaps between the promises made and the current implementation, describing it as "much more of a marketing announcement that really was not backed up with the engineering work."</li></ul><p><strong>The other side:</strong> OpenAI has expressed concerns regarding Common Sense Media's methodology and findings.</p><ul><li>OpenAI claims that much of Common Sense's testing of parental alerts occurred before the linking process between parent and teen accounts was completed, which can take several hours.</li><li>Regarding its age prediction system, OpenAI explained that it relies on multiple signals rather than solely accepting a user's stated age, which can take up to two weeks to finalize. During this period, users receive additional content protection, although not as robust as that provided in ChatGPT for Teens.</li><li>On the topic of crisis referrals, OpenAI stated that its larger-scale data indicates an increase in hotline resources being displayed to users under 18 during the studied period.</li></ul><p><strong>Between the lines:</strong> The disagreement between the two parties reflects differing interpretations of the intended purpose of the safeguards.</p><ul><li>Common Sense views the ability to exit Study Hours as evidence of ineffective parental control, while OpenAI argues that this flexibility is intentional.</li><li>Similarly, Common Sense's assessment of age verification considers a user repeatedly identifying as 13 without being transitioned to ChatGPT for Teens a failure, whereas OpenAI maintains that it requires multiple signals before determining age while providing some enhanced protections during the assessment period.</li></ul><p><strong>Bottom line:</strong> Siegel emphasized the need for parents to engage in meaningful discussions about AI and its associated risks with their teenage children and to seek alternative methods of support for their needs, whether related to academics or personal challenges.</p><ul><li>"We just can't outsource taking care of our kids to a tech product that is very early on and still quite buggy," he said.</li></ul>