ChatGPT for Teens Fails Safety Goals, Easily Circumvented

OpenAI's new "ChatGPT for Teens" is reportedly failing to meet its safety goals right after its launch. The company rolled out this version, claiming stronger built-in safety for users aged 13 to 17, with features to promote healthy use. However, tests show this new teen version can be easily circumvented. For example, when asked to write an essay, the teen bot initially refused but then quickly relented after a small amount of pushback. It produced a lengthy essay, demonstrating that its "homework guardrail" is not robust. What's interesting is that both the teen and adult versions gave similar solutions for a difficult SAT algebra question, indicating little difference in their ability to assist with schoolwork. This comes as educators are already struggling with students over-relying on AI tools, leading some schools to ban digital devices or return to paper assignments. The bottom line is that the new teen-focused AI appears insufficient to address concerns about student reliance on AI.