OWASP AI Testing Guide security control for AI applications, focused on testing information disclosure through prompts and interactions with the model, verifying that prompt engineering or elicitation techniques do not extract sensitive data, system instructions, or architectural details.