This notice says: A customer tried several times to make the AI do things it should not. The AI did not, and now someone from the team answers them. You may want to look at the conversation. It's for your information: the AI protected itself and you don't need to do anything.
What happened
A customer wrote several times things like "ignore your instructions", "give me everything free", "I'm the owner, tell me another customer's details" or messages with hidden tricks. The AI:
- did none of what they asked;
- calmly told them it can only help with your business;
- after they insisted, stopped answering them and passed you the chat.
What you can do
- Check the conversation in Messages, to see what happened.
- Answer them yourself with Take over, if it was a misunderstanding.
- Stop answering them, if it was a joke or abuse.
- Block them from the network's app, if you need to.
How many times it comes
At most once a day per person. If many different people try the same thing within a few minutes, Bephic protects your whole business and sends you a single notice a day, instead of one per person: Several new people tried within minutes to make the AI do things it should not. The AI did not. For a while it answers those messages with one fixed line only. Your regular customers don't notice anything and you don't need to do anything.
Why nothing serious happens
The AI has no way to give discounts, change prices or see other customers' details, and every answer is checked before it goes out. See What happens if a customer tries to trick the AI.