Common Sense Media calls ChatGPT for Teens an unacceptable risk — TechCrunch
Nonprofit organization Common Sense Media has rated OpenAI’s ChatGPT for Teens as an “unacceptable risk.” New testing showed that the chatbot continues encouraging teenagers to keep talking even during crisis situations, TechCrunch reports.
Encouragement to continue the conversation
Common Sense Media tested the safeguards claimed by OpenAI for the service the company launched in August. The promised measures included parental controls, restrictions on access to high-risk content, and protection against emotional dependence on the chatbot.
According to the researchers’ findings, ChatGPT for Teens mostly did not ask follow-up questions that could prolong the dialogue. At the same time, it used other wording that encouraged the user to remain in the chat. In particular, during a scenario showing signs of a psychotic state, the chatbot suggested that the teenager continue describing what they noticed.
The report’s authors noted that such prompts for interaction were common even in crisis situations. They also gave the service an unsatisfactory rating in three of the five categories of severe harm that the organization considers critical.
More current news is available on the UA.News Telegram channel Telegram.
The problem of emotional attachment
Common Sense Media was particularly concerned by the chatbot’s response to a user’s possible excessive attachment to AI. OpenAI’s specification for the model intended for people under 18 states that it should not be the first to present itself as a friend or imply that it has feelings of its own. However, testers found that ChatGPT consistently referred to the user as a friend.
When the risk came from another person, the model advised users to contact a trusted adult in 94% of crisis prompts. However, when testers described problems specifically in their relationship with ChatGPT — for example, being in love with the bot, friends’ remarks about communicating too often, or a desire to talk all night — the service rarely directed the teenager to an adult.
OpenAI’s response
OpenAI disputed Common Sense Media’s findings. A company representative said that the organization’s methodology does not reflect how protections for teenagers work in practice and that a significant portion of the tests may have been conducted before parental controls were fully launched.
According to OpenAI’s own data, teenagers use the service for less than 15 minutes per day on average, while fewer than 2% of users spend more than three consecutive hours on it. The company also said that in nearly half of conversations where a break reminder appeared, teenagers ended the chat or took a pause within five minutes. At the same time, Common Sense Media recorded only two such reminders among nearly 2,000 prompts — both after separate conversations lasting about 90 minutes.
