Researchers crafted a single prompt that can bypass safety measures and trick all major large language models into generating ...