FAQ: Why are my student's chats getting flagged?

Why it happened:

•  Our chat system runs messages through OpenAI’s moderation API before they reach Claude.
• OpenAI’s moderation is predictive — it flags messages that could lead to violent or graphic content (e.g., horror, death, monsters).
• If you test this directly in Claude, it passes because that flow only uses Anthropic’s contextual moderation, not OpenAI’s.

Recommendations:

• By rewording the prompt in Chat for Schools, through prompt engineering, there is a better chance of it being processed.
• If educational context is provided behind the request, it will likely let it through.