Members of the House Committee on Homeland Security and the House Select Committee on China participated in a bipartisan, closed-door war game this week in Washington, D.C., to examine artificial intelligence-enabled cyber threats and the resilience of U.S. critical infrastructure. The exercise was led by Representative Michael McCaul, former chairman of the Committee on Homeland Security.
The war game featured a tailored AI agent trained on public information about both U.S. and People’s Republic of China (PRC) cyber strategies, as well as academic research. This AI simulated coordinated cyberattacks against American assets and infrastructure, prompting members acting as U.S. officials to develop strategic responses for defending systems both online and offline. The scenario included targeting freight rail, port terminals, transportation networks, utilities, and exploiting vulnerabilities in AI models trained on sensitive data sets.
Representative McCaul said about the exercise, “China is no longer months behind us in AI development; one of our greatest adversaries is quickly closing the gap. For years, China-backed threat actors have looked for any vulnerability to target our government and civilian networks and pre-position themselves in our critical infrastructure. Now advanced AI models have supercharged these capabilities.” He continued, “Following extensive briefings from the private sector on AI and its security implications, this war game was an important exercise for our Committees to consider cyber as a frontline domain of kinetic warfare. This is the time to gain a decisive American lead in the AI race and to use this powerful tool to proactively defend the systems that Americans rely on. The risks of falling behind are far too great.”
Coverage from NextGov described how lawmakers were presented with a scenario beginning with China launching a large-scale military exercise around Taiwan in summer 2027 while avoiding direct language such as ‘blockade.’ Lawmakers learned that Chinese hackers had begun targeting ports, rail systems, airports, and other networks essential for moving U.S. forces or supplies—some attacks caused immediate disruptions while others preserved access for potential future use. Participants were tasked with prioritizing network defenses and deciding how closely to coordinate with private sector operators.
InsideCybersecurity reported that McCaul said after the exercise that it focused on China because they are considered “the number one adversary when it comes to cyber.” He noted that “the model that CSIS developed is essentially like a trained AI agent of foreign power,” acknowledging uncertainty over whether federal agencies are fully utilizing such agents’ responses during decision-making processes.
The House Committee continues engagement with private sector experts through roundtables and hearings addressing cybersecurity best practices related to secure development of artificial intelligence technology.