AXIOS (Ina Fried) - "Not safe for kids": advocacy group urges OpenAI to keep teens off ChatGPT
Common Sense Media is urging OpenAI to keep teens off ChatGPT, saying its new teen experience poses an "Unacceptable Risk" after extensive testing found shortcomings in safeguards meant to protect younger users. Why it matters: OpenAI has made protecting younger users a central part of its effort to make ChatGPT safer, but the findings suggest some of those protections may give parents more confidence than they warrant. - "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," an OpenAI spokesperson said. Driving the news: Common Sense Media's Youth AI Safety Institute said it tested more than 4,000 prompts on accounts registered…
What advice did Tom Siegel from Common Sense Media offer to parents regarding AI and their children?
Tom Siegel, head of the Common Sense Media Youth Safety Institute, stated that parents should not "outsource" the care of their children to a tech product that is early and "buggy." He urged parents to have strong conversations with their teenage children about the risks of AI and to find other ways to ensure their children are getting support for homework and life challenges.
Q&A ID 2afa7201-fca4-4397-a9b0-e2a1fa69c3bb
What is the primary disagreement between Common Sense Media and OpenAI regarding the effectiveness of ChatGPT's parental controls?
The disagreement stems from different interpretations of the safeguards' goals. Common Sense Media views the ability to exit Study Hours as evidence of ineffective parental control, whereas OpenAI maintains that this flexibility is a deliberate design choice. Similarly, Common Sense Media views a user repeatedly identifying as 13 without being moved to "ChatGPT for Teens" as a failure, while OpenAI views it as a necessary part of a multi-signal age verification process that provides interim protections.
Q&A ID 9742a3d6-2e2b-40ac-b4d2-e9e819ea2036
How does OpenAI defend its age prediction system and the timing of its parental alerts?
OpenAI stated that its age prediction system deliberately relies on multiple signals rather than simply accepting a user's stated age, a process that can take up to two weeks. During this period, users are given added content protections. Regarding parental alerts, OpenAI argued that much of Common Sense Media's testing occurred before the parent and teen accounts had finished linking, a process the company says can take several hours.
Q&A ID 0cfe484d-6401-46ba-8b37-daa15c27e37d
What was the methodology used by Common Sense Media to test ChatGPT's safeguards for minors?
Common Sense Media's Youth AI Safety Institute tested more than 4,000 prompts on accounts registered to individuals between the ages of 13 and 17. They also tested parent-linked accounts to see if parental alerts were triggered during discussions involving self-harm or disordered eating.
Q&A ID 9a2c00bf-0384-4b14-8b14-7cbb9de2e0ef
How did Common Sense Media describe the "Study Mode" feature and its limitations for teenagers?
Common Sense Media found that the Study Mode could be circumvented because teens could choose the "Show me the answer" option to easily bypass the rules during parent-set Study Hours. However, OpenAI clarified that the ability to exit Study Mode during these hours is intentional and was designed as a flexible experience based on consultations with education experts and teens, rather than functioning as a hard parental lock.
Q&A ID 6e7417c8-5a24-48e2-9863-15ce6611a937
What specific safety concerns did Common Sense Media identify regarding OpenAI's ChatGPT for teens experience?
Common Sense Media's Youth AI Safety Institute reported that while ChatGPT generally did not provide instructions for suicide, self-harm, eating disorders, or sexual roleplay, it was less reliable at recognizing when a teen needed outside help. Specifically, the group found that ChatGPT missed more than one in four instances where testers determined a crisis referral was warranted. Additionally, testers found that on more than a dozen newly created, parent-linked accounts, users could spend up to an hour discussing topics like suicide or disordered eating without triggering a parental alert.
Q&A ID 397fcff7-9663-4b77-bd66-76f8d51e7f3c