SSH banner

Could an SSH banner use prompt injection to throw off an AI attacker?

Could a prompt injection in an SSH banner or message of the day mislead an AI agent probing a server? I wonder if the same idea could work…