The NGO believes that the teenage version of ChatGPT fails to alert parents when it should and always does their homework; OpenAI questions the testing method.
Common Sense Media, an American association specializing in the evaluation of digital services for young people, classified ChatGPT for Teens as an "unacceptable risk". Launched in August, this version of ChatGPT for teenagers promised reinforced safeguards and a use focused on learning. According to the NGO's assessment, it does not keep these promises: notifications to parents do not go off when they should, the help offered in crisis situations is insufficient, and the tool continues to do homework for students.
Tom Siegel, who heads Common Sense Media's institute dedicated to youth safety from AI, believes that these safeguards can give parents a deceptive confidence, as a teenager can talk about self-harm for an hour without any alerts going off. He recommends reserving ChatGPT for adults until OpenAI has corrected the problem and proven the effectiveness of its protections through independent testing.
OpenAI challenges
OpenAI claims that most of the tests may have been conducted before the activation of parental controls was effective, which would distort the results on notifications. The NGO maintains its methodology: the company only reported after the fact that activation could take several hours, and some test accounts, linked for much longer, did not receive an alert either.
Read the full article on the original source
Go to the source →