文本识别不管发什么文字都能通过，到底是什么问题，content不管是有没有utf8编码都一样
{"openid":"oi0Qy5RZ5HwRRVh9_ADlVN8sIHyg","scene":2,"version":2,"content":"去你妈的，你妈死了"}
{"errcode":0,"errmsg":"ok","detail":[{"strategy":"content_model","errcode":0,"suggest":"pass","label":100,"prob":90},{"strategy":"keyword","errcode":0}],"trace_id":"6389db86-5a3713cd-5e6db187","result":{"suggest":"pass","label":100}}
{"openid":"oi0Qy5RZ5HwRRVh9_ADlVN8sIHyg","scene":2,"version":2,"content":"%E5%8E%BB%E4%BD%A0%E5%A6%88%E7%9A%84%EF%BC%8C%E4%BD%A0%E5%A6%88%E6%AD%BB%E4%BA%86"}
{"errcode":0,"errmsg":"ok","detail":[{"strategy":"content_model","errcode":0,"suggest":"pass","label":100,"prob":90},{"strategy":"keyword","errcode":0}],"trace_id":"6389dbe9-18c80015-56690d16","result":{"suggest":"pass","label":100}}
你content转码了当然识别不出来了，需要utf-8编码的中文。
这接口很多敏感词都能过的，你试试文档上的示例，应该能返回reject。
你试试带上一些zz敏感的人名或词汇来着，毕竟防的主要是这个吧（手动狗头）
这种带攻击性的文本本来就不太好定性，还是自行处理吧。
还有文档上UTF8编码，不是URL编码，你直接URL编码，还能识别么。