U.S. Lacks Strategy to Manage AI as National Security Threat
The United States is losing ground in its competition with authoritarian powers over artificial intelligence development, and the government has no coherent plan to protect critical infrastructure from AI-enabled attacks. Two former officials who shaped AI policy under different administrations say Washington must act now or risk being overwhelmed by the technology's rapid advancement.
AI systems have advanced dramatically in recent years. What once struggled to write coherent sentences now scores above expert humans on a wide range of tests, and the gap continues to widen as more powerful systems help design even more capable successors.
Immediate Threats to Infrastructure
Recent developments show the scale of the risk. Anthropic's Claude Mythos Preview model can detect subtle coding errors and has identified thousands of critical vulnerabilities in foundational software that runs computers and the internet. Some vulnerabilities had gone undetected for decades.
In hostile hands, such AI tools would enable attacks on power grids, hospital systems, and banking infrastructure across the country. OpenAI's GPT-5.4 model now outperforms Ph.D.-level virologists at troubleshooting lab work, while Mythos matches top human experts in capabilities needed to develop bioweapons.
AI systems are also advancing in materials science, software development, and industrial processes-all essential to designing new weapons. In Ukraine, AI is making weapons systems themselves more autonomous.
A Bipartisan Problem Requiring Bipartisan Solutions
Both parties recognize the danger. Officials from different administrations agree that clear, actionable steps exist that could command bipartisan support. Yet Washington has not acted with the necessary urgency.
Without a strategic shift, AI development will outpace the government's ability to manage it. A coordinated national approach to AI security is achievable-but only if leaders prioritize it now.
Learn more about AI for Government and AI for Cybersecurity Analysts to understand how these technologies affect your work.
Your membership also unlocks:
