finishReason `BLOCKLIST`? How to allow "forbidden terms" in the ouput

Gemini API
,
1

I’m getting the BLOCKLIST finish reason when asking Gemini 1.5 pro to transcribe legal documents that deal with sensitive issues. The gemini docs say this is b/c “Token generation stopped because the content contains forbidden terms.”.
Questions:

  1. How can I can turn this off? I have already turned off all the available safety settings in the api.
  2. Can I ask the model to just skip the forbidden term (masking it with **** maybe?), that would be OK.
  3. Is this list of forbidden terms available anywhere so I can consult and try to work around it?

Thanks.

Docs reference: Generating content  |  Gemini API  |  Google AI for Developers

2

System prompting the model to censor explicit/offensive language could resolve the issue, because the “BLOCKLIST” is being returned in finishReason meaning it’s the model generated content that’s causing the issue.