This research evaluates the safety of Large Language Models when interacting with children, identifying unique vulnerabilities and proposing targeted safeguards.

• Identifies safety gaps in current LLMs specific to children's developmental stages
• Examines transformative applications in education and therapy alongside potential risks
• Proposes a comprehensive evaluation framework that accounts for children's diverse needs
• Addresses the often-overlooked differences in safety requirements across age groups

Why it matters: As LLMs become integrated into educational tools and therapeutic applications for minors, understanding and mitigating child-specific security risks is essential for responsible AI deployment.

LLM Safety for Children