GPT-Red
Named by 5 kept items in this workspace.
2 items name it under frameworks and tools, 2 under models and 1 under techniques.
The items that name it 5
Automated red-teaming agent and automated evaluations used for adversarial training and for measuring direct and indirect prompt-injection robustness.
Automated red-teaming approach used to evaluate direct and indirect prompt injection, described in the GPT-6 Astra system card.
In every workspace you can read 12
Security
All technologies in this workspace.