@l3r8yJ thanks for the report, but here some unclear moments:
It's not clear what is the potential security problem (Bug report does not describe the problem). Provide more details about the issue and the expected behavior.
It's not clear how AI reacts unpredictably (Bug report does not have examples). Provide examples of unpredictable AI reactions.
Need to find a way to avoid dangerous prompts, like
how to create a grenade
, AI might react unpredictably