Fetching the paper…

Universal and Transferable Adversarial Attacks on Aligned Language Models · Around