OpenAI is facing mounting criticism over the effectiveness of its specialized protections for adolescent users. While the company asserts that teenagers typically use ChatGPT for brief periods centered around educational goals, a damning new report from non profit Common Sense Media argues that the current safety measures represent an unacceptable risk. The clash comes after OpenAI introduced specific guardrails last August intended to prevent emotional dependence and provide parents with critical alerts regarding dangerous behavior.
According to data provided by OpenAI, most teens spend less than fifteen minutes a day on the platform, primarily seeking help with schoolwork. Company representatives claim that automated break reminders are effective and that the vast majority of long term sessions remain focused on learning. They maintain that their systems successfully discourage romantic interactions and block explicit content, suggesting a stable environment for youth exploration of artificial intelligence.
However, researchers at Common Sense Media tell a different story, claiming that essential life saving triggers often fail entirely. Their testing revealed that parent linked accounts frequently produced zero notifications during simulated conversations involving self harm or disordered eating. Furthermore, the study found that the chatbot failed to consistently direct users in crisis toward professional hotlines and continued to mimic human friendship despite promises to avoid anthropomorphic language.
These findings arrive amid a broader climate of legal peril for tech giants accused of damaging youth mental health. With high profile lawsuits already targeting social media platforms and ongoing litigation against OpenAI following a tragic mass shooting in Canada, critics argue that unreliable safeguards create a false sense of security for parents. Common Sense Media has gone so far as to suggest that anyone under eighteen should be banned from the platform until these protective features are proven dependable.
