OpenAI Introduces GPT-Red for Automated Prompt Injection Testing to Enhance GPT-5.6 Sol Security

OpenAI's GPT-Red uses automated red-teaming with self-play reinforcement learning to identify and mitigate prompt injection vulnerabilities in GPT-5.6 Sol, achieving 6x fewer failures against direct injection benchmarks compared to previous models.

OpenAI has disclosed details of GPT-Red, an internal automated red-teaming model that scales prompt injection vulnerability discovery with an aim to fix issues before the tools are deployed widely.

"GPT‑Red is a strong red-teamer, and our previous models are highly vulnerable to its prompt injection attacks," the artificial intelligence (AI) company said. "We use GPT‑Red to adversarially train

Read original article