Threat Modeling

AI Red Teams apply threat modeling to identify unique attack surfaces in AI systems, such as manipulating training data, exploiting prompt interfaces, attacking the model inference process, or compromising connected tools/APIs. Before attacking an AI system, red teamers perform threat modeling to map out possible adversaries (from curious users to state actors) and attack vectors, prioritizing tests based on likely impact and adversary capability.

Learn more from the following resources:

• @article@Core Components of AI Red Team Exercises (Learn Prompting)
• @guide@Threat Modeling Process
• @guide@Threat Modeling