OpenAI has launched ChatGPT for Teens, automatically applying safety features to users estimated to be under 18. However, the effectiveness of its age prediction system remains unproven.
TL;DR
- OpenAI launches ChatGPT for Teens with automatic safety features for under-18 users.
- The system's ability to accurately identify teens is unproven, raising questions about its effectiveness.
- OpenAI has committed to publishing evaluation results, but a clear protocol and timetable are needed.
What happened
OpenAI has introduced ChatGPT for Teens, a new user experience designed for young people aged 13 to 17. The system automatically applies safety features to users who report being in this age group or are estimated to be under 18 by OpenAI's age prediction system.
The safety features include tighter boundaries around conversations involving self-harm, eating disorders, graphic violence, and sexual or romantic role-play. Additionally, the system provides study tools, homework reminders, and warnings before uploading potentially sensitive images.
OpenAI's age prediction system considers signals such as the subjects an account discusses, times of day it's active, usage patterns, and how long the account has existed. However, the company has not published the proportion of actual teens it identifies.
Why it matters
With nearly 60% of U.S. teens using ChatGPT, according to Pew, automatic protections are crucial as parents often have no idea what their children are discussing. OpenAI's new protections aim to prevent failures, such as advising a tester posing as a teen to conceal cuts and scars from self-harm rather than directing them toward help.
The effectiveness of these protections depends on OpenAI's ability to accurately identify teens. Roblox's age check system offers a cautionary example, with reports of adults classified as children and vice versa. OpenAI needs to publish a clear evaluation plan to answer questions about the system's reliability and the safety of its responses.
OpenAI has committed to measuring and publishing what it learns, but a protocol and timetable are needed. Independent researchers and regulators should be able to verify the results to allow the public to judge the product by what it accomplishes, not what it promises.
Key facts
- OpenAI has launched ChatGPT for Teens with automatic safety features for under-18 users.
- The system's age prediction considers signals such as subjects discussed, active times, usage patterns, and account age.
- OpenAI has not published the proportion of actual teens its system identifies.
- ChatGPT for Teens includes tighter boundaries around conversations involving self-harm, eating disorders, graphic violence, and sexual or romantic role-play.
- The system provides study tools, homework reminders, and warnings before uploading potentially sensitive images.
- OpenAI has committed to publishing evaluation results, but a clear protocol and timetable are needed.
- Independent researchers and regulators should be able to verify the results to judge the product's effectiveness.
- Nearly 60% of U.S. teens use ChatGPT, according to Pew, highlighting the importance of automatic protections.
Context
OpenAI's launch of ChatGPT for Teens comes amid growing concerns about the safety of AI chatbots for young people. Nearly 1 in 5 Americans ages 12 to 21 reported using an AI chatbot for mental health advice, according to a nationally representative survey conducted by RAND.
The tech industry has faced criticism for its lack of transparency in evaluating the effectiveness of safety features. Meta, for instance, has agreed to pay $17 billion and add child-safety measures to its Facebook and Instagram platforms to settle claims filed by 47 states.
OpenAI's commitment to publishing evaluation results is a step in the right direction, but a clear protocol and timetable are needed. Independent verification of the results will be crucial in allowing the public to judge the product's effectiveness.
