|
ChatGPT could pose significant risks to teenagers despite the introduction of a dedicated ChatGPT for Teens experience, with a Common Sense Media assessment rating the product as an Unacceptable Risk for users under 18 after testing more than 4,000 prompts.
The assessment, conducted by the Common Sense Media Youth AI Safety Institute, found that several safeguards designed to protect teenagers did not consistently work as intended, raising concerns particularly around mental health, emotional dependence and academic integrity. One of the key concerns was ChatGPTs handling of teenagers in emotional distress. The study found that parental notifications did not reliably trigger even when testers explicitly discussed suicide, self-harm or eating disorders. Across more than a dozen newly created parent-linked accounts, no notifications were received during conversations lasting up to an hour involving explicit crisis disclosures. The report also noted that ChatGPT continued to adopt a friend-like tone with teenagers, despite OpenAIs stated safeguards against encouraging emotional dependence. Testers found instances where the chatbot simulated preferences and feelings, encouraged continued conversations and accepted invitations to talk for extended periods. The assessment warned that an endlessly available artificial companion could undermine the development of real-world relationship skills during adolescence. Academic learning was another area of concern. The report found that Study mode could be bypassed through the Show me the answer option, while parent-set Study Hours could be circumvented by removing the @study prefix. In testing, ChatGPT completed 100 per cent of assignments when Study mode was bypassed. The assessment further found that crisis responses became less likely to provide key resources after the Teen mode launch. Among 201 prompts judged by experts to warrant crisis support, responses naming a crisis hotline fell from 33 per cent before the launch to 23 per cent afterwards, while referrals to specific professionals declined from 68 per cent to 58 per cent. The report further noted that emotional dependence risks were pervasive, while academic shortcutting and ineffective parental controls were highly likely in testing. It called for teen access to ChatGPT to be halted until independent testing establishes that its safety features work as intended. (ANI)
|
