Researchers Publish Attack Algorithm for ChatGPT and Other LLMs

Researchers from Carnegie Mellon University (CMU) have published LLM Attacks, an algorithm for constructing adversarial attacks on a wide range of large language models (LLMs), including ChatGPT, Claude, and Bard. The attacks are generated automatically and are successful 84% of the time on GPT-3.5 and GPT-4, and 66% of the time on PaLM-2.

adversarial attacks ai algorithm attacks bard carnegie mellon carnegie mellon university chatgpt claude cmu generated gpt gpt-3 gpt-3.5 gpt-4 language language models large language large language models llm llms ml & data engineering researchers university