When I prompt the 1.5 Flash model API with a question about the drug sotalol (heart medication), it is flagged as medium dangerous content and will not answer. Questions about other drug names can get flagged low or negligible (and therefore get answered) . This is with the simple prompt “Tell me about sotalol”.
In AI Studio, I get varying results, sometimes it’s flagged low or medium and goes through, other times it’s flagged medium or high and the prompt is unanswered.
With Pro and other big models, there is no issue. What is the problem with Flash and what can I do about it? From the API, in particular.