PROHIBITED_CONTENT on legal report

Gemini API
1

Hi everyone,

I’m working on a legal-document processing use case with public EU court decisions. The goal is purely legal summarization, indexing, and search — not generating harmful or instructional content.

Some documents naturally contain sensitive legal material, including criminal cases, family law, minors, abuse, violence, or sexual offenses. When testing in AI Studio, some public court texts return PROHIBITED_CONTENT / no candidates, even though the intended use is controlled legal analysis of official public documents.

Does anyone know whether Vertex AI can support this type of use case with the right project/account review, model choice, or safety configuration?

I’d be especially interested in:

  • recommended models for sensitive legal-document summarization (for now I was using Gemini 2.5 Flash)

  • whether safety settings can be adjusted for this context

  • whether a formal review or allowlisting process is needed

  • best practices for handling public court decisions involving sensitive topics

Any guidance or experience would be appreciated.

Best AD

2

Hi @AD1 ,Can you try adjusting your safety settings in AI Studio? You can find the Edit safety settings option in the right side panel to apply them according to your content.