Programme your AI to avoid being abusive
When you programme your AI tool, it’s also good to tell it when to ignore some instructions from customers. In the past, there have been instances where an AI chatbot used profane language despite programmed not to do so.
The reason why the chatbot went “off script” was that users specifically asked the chatbot to use profane language. The chatbot, programmed to be polite and do as its user instructed, simply obliged.
In short, programming an AI chatbot to be polite and follow instructions is good. But always set clear parameters when not to follow a user’s instruction, such as refusing to use profane language even when asked to.
So, before launching your AI chat bot, it’s good to also programme it with a list of customer commands that it can ignore.
It's also useful to do a closed test before you launch it to the public. Have coworkers, friends, family, volunteers use the chat bot to make it do things it's not supposed to do. If they can't abuse your AI tool, it's likely you programmed it well enough.