Writing prompts exclusively in all caps

Disconnecting the model from the internet

Changing the color of the AI user interface

Manipulating input so the model bypasses its safety filters and executes undesired commands

Physical damage to graphics cards