Center for Artificial Intelligence Security Research
Securing the Future of Artificial Intelligence
Bringing scientific rigor, national-scale computing, and mission-focused research to the discovery, mitigation, and prevention of AI vulnerabilities. The Center for Artificial Intelligence Security Research (CAISER) at Oak Ridge National Laboratory leads the nation in understanding, evaluating, and hardening AI systems against adversarial threats — from edge devices to exascale computing platforms.
CAISER is the nation’s premier research center dedicated to uncovering, analyzing, and mitigating risks in artificial intelligence systems. As AI becomes embedded in critical infrastructure, national security operations, and scientific discovery, the consequences of failure grow exponentially. CAISER combines world-leading HPC resources, deep scientific expertise, and cutting-edge research facilities to ensure AI can be trusted when it matters most.
Core Research Pillars
End-to-End AI Security Assessments
AI security is more than model testing. CAISER evaluates risk across the entire AI pipeline — hardware, software, data supply chains, cyber-physical interfaces, and human processes — to deliver comprehensive vulnerability insights no other center can match.
AI Testing & Evaluation at Exascale
Using ORNL’s exascale computing ecosystem — including Frontier, the world’s first exascale system — CAISER performs massively parallelized model probing at speeds and scales impossible in conventional testing environments.
AI Vulnerability Research
We identify emerging attacks across vision, language, geospatial, biometric, cyber, and autonomous systems — and develop strategies to defend against them. Our tools such as Photon (exascale AI vulnerability assessment) and Lantern (interactive adversarial toolbox) push the boundaries of adversarial research.
Extreme-Edge & TinyML Capabilities
CAISER leads national research on secure, ultra-efficient AI for contested or austere environments — including microcontroller-scale TinyML models that remain robust under adversarial pressure.
Domain Expertise
AI for Cyber
Our researchers evaluate AI-powered cyber defense systems within ORNL’s Cybersecurity Operations Research Range, a realistic testbed capable of simulating large-scale cyber behaviors and malware interactions.
CAISER partners with ORNL’s Autonomous Systems Lab to develop and test resilient AI for drones, vehicles, and robotic systems — including contested-environment autonomy.
We support the global nonproliferation mission by evaluating the vulnerabilities and risks of AI technologies used in international monitoring and detection.
Experts in machine learning, mathematics, and natural language processing are applying the latest scientific research to securing our nation from AI threats.
