Dual-Use Dilemma

A core concept in AI safety referring to the situation where the same open-source model weights or technology can be used for legitimate research and also exploited by malicious actors for harmful purposes, with no technical means to isolate the two uses from each other.