試験
通常
試験時間: 00:00問題時間: 00:00
問題一覧
問題 33 / 65
In an AI analytics system for government agencies, if confidential information has been embedded in the system prompt, which countermeasure is most effective?
①Completely remove confidential information from the system prompt and securely manage it on the server side.
②Frequently embedding confidential information into prompts for convenience.
③Operate it as-is, assuming the model won't automatically output confidential information.
④Encrypt all communications with TLS before sending the prompt.
⑤Encrypt the prompt before sending it, then decrypt it on the receiving end and use it as-is.