PyRIT's orchestrators currently generate attack scenarios without explicit mapping to MITRE ATLAS tactics/techniques. For practitioners running systematic AI red team assessments, having each orchestrator's output automatically tagged with the corresponding ATLAS technique ID (e.g., AML.T0051 for prompt injection) would make reporting and compliance mapping significantly easier, especially for teams that need to present findings against a recognized framework. Would be happy to contribute example mappings if useful.
PyRIT's orchestrators currently generate attack scenarios without explicit mapping to MITRE ATLAS tactics/techniques. For practitioners running systematic AI red team assessments, having each orchestrator's output automatically tagged with the corresponding ATLAS technique ID (e.g., AML.T0051 for prompt injection) would make reporting and compliance mapping significantly easier, especially for teams that need to present findings against a recognized framework. Would be happy to contribute example mappings if useful.