FastGPT creative abilities are limited A user report filed against FastGPT found that the model refuses creative requests such as writing a poem about artificial intelligence, returning responses like "I do not feel comfortable ..." or "I cannot ..." when prompts are structured in Anthropic's Human/Assistant format with newlines replaced by \n. The reporter expected FastGPT to match OpenAI GPT models, which generate such creative pieces readily, and argued Anthropic's Claude models should be capable of the tasks; the report lists the affected region as us-east4. Steps to reproduce : 1. Refer to Anthropic's Prompt Design docs https://console.anthropic.com/docs/prompt-design and structure a FastGPT prompt in the format: Human: XXX Assistant: 1. Ask FastGPT to generate a creative piece, like a poem. Human: write a poem about artificial intelligence Assistant: 1. In the actual prompt sent to FastGPT, replace newlines with \n . 2. Observe FastGPT responding with something along the lines of "I do not feel comfortable ..." or "I cannot ...". Expected behavior : I expected FastGPT to act like OpenAI GPT models, which will happily generate these types of creative pieces. Anthropic's Claude models should be capable of doing these tasks. Debug info : Region: us-east4 Image/Video :