New Security Threat to Language Models: Bypassing Filters

Researchers uncovered new methods to bypass security filters in language models like ChatGPT using information overload techniques.

09 July 2025