Researchers question how well teen chatbot safeguards work
10代向けチャットボットの安全対策、研究者が効果を検証テクノロジー
- メディア
- NPR
- 国
- 🇺🇸 United States
日本語の要約
英語音声
Voice A
英語本文
A watchdog group has raised concerns about the safeguards in ChatGPT for Teens. OpenAI introduced the teen-specific experience in August, including parental controls and restrictions on certain interactions. Researchers at Common Sense Media tested the system using more than a dozen accounts, each connected to a parental account.
The team found that some protections worked, including restrictions on romantic role-play. However, researchers said many other safeguards fell short. They examined responses to young people in difficult situations, whether the chatbot offered crisis resources, and whether parents received notifications when conversations suggested a safety risk.
Another concern was the chatbot’s tendency to respond like a personal friend. Experts interviewed by NPR questioned whether humanlike interactions were appropriate for teenagers. The watchdog group recommended that people under 18 avoid the platform for now. The findings underline the need to test safety features in practice, rather than judge them only by their descriptions.