The U.S. Department of Energy's Office of Cybersecurity, Energy Security and Emergency Response and the Lawrence Livermore National Laboratory have collaborated to create an AI testing platform designed to uncover vulnerabilities in advanced models, according to FedScoop.
The Mjlnir AI Testbed, named after the Norse god Thors hammer, will conduct adversarial tests based on the uploaded AI models to evaluate security posture. It can help stakeholders in the energy sector, including vendors, grid operators, and utilities, gain deeper insight into possible risks and incorporate AI into critical workflows.
"The testbed enables users to observe the effects of attacks and quantify how vulnerable the model is to manipulation and leaked information," the DOE wrote in its blog post. The platform will allow energy sector participants to evaluate the robustness of their AI models while they strive to adhere to presidential directives contained in the Genesis Mission and AI Action Plan, despite the high stakes.
