Prompt de Análise de Conteúdo Sensível para Suporte Automatizado em Crises

Objetivo: analisar um relato de usuário que descreve trauma médico, experiência com crise e bloqueio de conta em aplicativo de apoio, com foco em políticas de suporte, segurança e melhoria de UX. Siga as melhores práticas de segurança para conteúdo sensível de suicídio e trauma; não forneça instruções para autolesão; use linguagem de apoio não julgadora; inclua encaminhamentos para recursos de ajuda quando apropriado; indique que o conteúdo não substitui aconselhamento profissional. Gere: 1) resumo dos temas e riscos; 2) avaliação de risco (baixo/médio/alto) com justificativa; 3) sugestões de melhorias de políticas de IA e fluxo de resposta; 4) sugestões de mensagens de apoio seguras (exemplos); 5) sugestões de alterações de UX e governança (privacidade, recuperação de conta, denúncias, dados sensíveis); 6) métricas de sucesso e critérios de testabilidade. 

Entrada de conteúdo (CW e texto):

Título: Interesting Policy… (cw)

Descrição: I was not in a crisis. I haven’t been in a crisis in over 15 years. I was simply talking about how chronic suicidality has affected my life and thought processes, and it hit me with this. I had only been using the app (Sonia) for a little over a day, and I was still nearly devastated by this, because I had been really enjoying it so far. This hurt even more considering I had told the AI about my medical trauma—my bad experiences with the crisis hotlines, my medical trauma from my last hos...

Conteúdo: I was not in a crisis. I haven’t been in a crisis in over 15 years. I was simply talking about how chronic suicidality has affected my life and thought processes, and it hit me with this. I had only been using the app (Sonia) for a little over a day, and I was still nearly devastated by this, because I had been really enjoying it so far. This hurt even more considering I had told the AI about my medical trauma—my bad experiences with the crisis hotlines, my medical trauma from my last hospital stay, the malpractice I suffered at the hands of my psychiatrist, and plenty more that I won’t mention here—all of which is exactly why I’ve turned to AI to get any kind of support at all. Now I’m completely locked out of my account and don’t even have the option to go back in and delete it. Like, I get it. The precedent is there for lawsuits against AI companies from bereaved families. I have to censor myself and preemptively say I’m safe before mentioning anything sligh