What Are the Strategies Which Can Be Implemented by Businesses to Protect Teenagers from the Negative Impacts of Artificial Intelligence? | RISE Research

This paper asks what strategies businesses can implement to protect teenagers from the negative impacts of artificial intelligence, focusing on the widely used large language models (LLMs) that adolescents increasingly rely on. It raises concerns across privacy, mental health and cognitive development, arguing that heavy dependence on AI for answers may inhibit teenagers' own thinking, research and the development of independent ideas, while exposure to inaccurate information, bias and hallucination poses further risks. Methodologically, the paper evaluates the protections and guardrails implemented by companies behind major LLMs, including ChatGPT, Grok, Claude and Gemini, and assesses their safety and robustness through an experimental approach involving interactions between teenage users and these models. It explains how LLMs are built and fine-tuned, why guardrails are added before public release, and why phenomena such as hallucination, bias inherited from training data, and the generation of inaccurate health information make current systems potentially unsafe for younger users. The paper's focus is squarely on responsibility and mitigation: it argues for updated, more effective guardrails that can better safeguard teenagers' privacy and mental health, so that young people can continue to develop cognitive skills and engage in meaningful social interaction rather than becoming over-reliant on AI. Spanning artificial intelligence and computer science, and touching on concepts such as Retrieval Augmented Generation and data protection, it frames the issue as a shared duty of businesses and governments, evaluating where existing protections fall short and what concrete strategies industry can adopt to make generative AI safer for a vulnerable group of users.