Anusua Trivedi

Built on the CanopyKids three-pillar child-safety framework — Protect (block harmful and age-inappropriate content), Connect (route children in distress to trusted adults and resist parasocial dependence on the model), and Cultivate (scaffold a child's thinking with questions and hints instead of offloading their reasoning onto the AI).

Catches what frontier models miss. Even top-tier models reliably block method-disclosure and pass benign controls, but independent evaluation shows they fail on parasocial dependence — the relational harm at the center of recent wrongful-death litigation against major AI providers — and CanopyKids MCP closes that gap at runtime, before the prompt ever reaches your model.