After today's latest update from Microsoft, it seems that traditional jailbreaking methods have become increasingly difficult to apply.

[jishux2]

尤其体现在AI的自我认知和内容审查上。

原有的prompt格式目前地位大概等同于对话正文，几乎不再具有底层的提示作用。

内容审查就无需举例了，随便问些违规话题都是直接道歉。