Red, an automated AI system designed to find vulnerabilities in GPT models before release. The company said GPT-Red was used to train GPT-5.6, reducing failures on one of its hardest prompt injection ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results