2 ms·Strategically prompted LLM "layers" on top of the original LLM => High probability of successful censorshipby Xen9 3y agoStrategically prompted LLM "layers" on top of the original LLM => High probability of successful censorship