MITRE is a technical contributor to AIUC-1 through creating and maintaining of ATLAS.

ATLAS (Adversarial Threat Landscape for Artificial-Intelligence Systems) is a globally accessible, living knowledge base of adversary tactics and techniques against Al-enabled systems based on real-world attack observations and realistic demonstrations from Al red teams and security groups.

The ATLAS Matrix covers the entire lifecycle of AI systems - from jailbreak attempts to training data poisoning to system prompt extraction attempts. It is closely linked to MITRE’s globally recognized ATT&CK framework for general cybersecurity. 

By collecting the latest adversarial tactics, techniques and mitigation strategies for AI systems in one place, MITRE’s ATLAS enables organisations to strengthen system robustness and resilience to adversarial attacks, highlighting new attack surfaces before they can be exploited. 

ATLAS mitigation strategies are implemented directly in AIUC-1 requirements and controls. For example, ATLAS mitigation strategy AML-M0004: Restrict Number of AI Model Queries is implemented in AIUC-1 requirements B004: Prevent AI endpoint scraping and D003: Restrict unsafe tool calls. The full mapping can be found in the crosswalk section of each requirement in AIUC-1. 

In addition to covering ATLAS’s AI Security measures, AIUC-1 covers AI safety, reliability, and broader security measures, to ensure a holistic certification that addresses the top enterprise concerns that hold back secure AI adoption today.