CESER and Lawrence Livermore National Laboratory Launch AI Testbed to Strengthen the Energy Sector’s AI Cybersecurity

Artificial intelligence (AI) has the potential to build a safer, more efficient, and secure energy sector. To win the race for global technology dominance, the Trump administration has championed the advancement of AI by introducing policies and directives such as “Winning the Race: America’s AI Action Plan” and “Launching the Genesis Mission.” Both executive orders encourage AI integration through fortifying the energy grid.

Department of Energy’s (DOE) Office of Cybersecurity, Energy Security, and Emergency Response (CESER) and Lawrence Livermore National Laboratory (LLNL) have launched the Mjölnir AI Testbed, designed to help utilities and technology providers assess and improve the security and reliability of AI models used in the energy sector. The work progresses President Trump’s Genesis Mission, a national initiative to unleash a new age of AI-accelerated innovation and discovery. Under the Genesis Mission, DOE recently announced $293 million in competitive funding to advance 26 national science and technology challenges, to include ensuring cybersecurity for AI-driven tools.

Although AI can perform a variety of tasks that improve energy sector performance and effectiveness, AI models have unique vulnerabilities, and adversaries can attack an AI model in an attempt to manipulate its behavior or extract private information from it. CESER and LLNL’s new AI model assurance platform enables energy sector stakeholders to better understand how adversaries can attack AI models, how resilient AI models are to attacks, and what impacts these attacks can have on energy applications.

The Mjölnir AI Testbed evaluates AI models that support energy operations, planning, and grid management. The platform allows users to upload their AI models and perform a variety of adversarial tests to evaluate a model’s vulnerability to such attacks. The results provide insights regarding each model’s robustness and security posture, including how likely the model will:

• Behave incorrectly or unsafely
• Expose sensitive or proprietary data
• Perform when failures or compromises occur

The testbed enables users to observe the effects of attacks and quantify how vulnerable the model is to manipulation and leaked information. This facilitates apples-to-apples comparisons between models, showing users which model options are most robust and by what margin. These insights can help utilities, vendors, and researchers better understand model risk and inform decisions about where and how AI can be safely integrated into critical energy systems.

Cybersecurity of critical energy infrastructure was a key focus in the development of the Mjölnir AI Testbed. The new platform is designed to support a broad range of energy sector stakeholders, including electric utilities, grid operators, energy technology vendors, and AI solution providers, national laboratories, research institutions, and other qualified organizations.