OpenAIs GPTRed Automates Prompt Injection Testing to Harden GPT56 Sol

OpenAI has disclosed details of GPTRed an internal automated redteaming model that scales prompt injection vulnerability discovery with an aim to fix issues before the tools are deployed widely GPTRed is a strong redteamer and our previous models are highly vulnerable to its prompt injection attacks the artificial intelligence AI company said We use GPTRed to adversarially train