Report Urges Federal Government to Designate Frontier AI Models and Data Centers as Critical Infrastructure

Frontier artificial intelligence models are fundamentally altering the balance of power in cybersecurity, accelerating the discovery of software vulnerabilities to machine speed. For years, critical infrastructure operators relied on the assumption that they could patch systems faster than adversaries could operationalize exploits. That fragile equilibrium is now collapsing, as AI-generated scripts shrink the timeline between vulnerability discovery and weaponization.

In response to this shifting landscape, a growing coalition of policy and security experts is arguing that the infrastructure powering these advanced models can no longer be treated as ordinary commercial assets. Instead, they argue, the systems that train and deploy frontier AI have become so foundational to the digital economy that they require federal protection.[3][4]

A new report published Thursday by the nonprofit Americans for Responsible Innovation (ARI) proposes a structural shift in U.S. cyber defense, urging the federal government to officially designate the AI sector as critical infrastructure. The proposal seeks to formally recognize that the rapid adoption of large language models across both the private sector and federal agencies has created a new, highly concentrated attack surface.

According to the report's authors, the AI ecosystem is now deeply interwoven with essential public and private services, meaning that a single successful attack on foundational AI systems could trigger cascading failures across multiple industries simultaneously.[1]

The ARI report defines the AI sector broadly, encompassing far more than just the software companies developing the models. The proposed critical infrastructure designation would cover frontier model designs, model weights, and alignment systems, alongside the physical architecture that makes them possible.

This includes the massive data centers housing specialized compute clusters, the semiconductor chips powering them, and the broader platforms used to deploy AI models at scale. By grouping these elements into a single critical infrastructure sector, policymakers could prioritize limited federal resources to protect the entire AI supply chain from disruption.[1][7]

If adopted, the designation would unlock a wide range of federal tools and resources for AI companies, many of which are currently reserved for sectors like energy, water, and telecommunications. These resources include operational continuity planning, incident response services, and access to bespoke, real-time threat intelligence. Furthermore, it would allow AI developers and data center operators to integrate with federal systems like the Continuous Diagnostics and Mitigation (CDM) program, providing a layer of proactive defense against both state-sponsored adversaries and sophisticated cybercriminal syndicates.[1]

To manage this new sector, the report calls for naming the Cybersecurity and Infrastructure Security Agency (CISA) as the lead federal agency. CISA already serves as the national coordinator for critical infrastructure security, managing 16 existing sectors and operating initiatives designed to help operators sustain essential services during severe cyberattacks. Programs like CISA's CI Fortify initiative, which helps critical infrastructure operators proactively isolate networks and maintain operations during crises, could be extended to protect the data centers and training environments where frontier models are developed.[2][4]

The push for federal protection comes amid escalating, real-world threats to both the physical and digital components of the AI ecosystem. Data centers, which are costly to conceal and consume vast amounts of electricity and water, have increasingly become visible targets in geopolitical conflicts.

Disabling key physical components, such as cooling systems or power transformers, can take large computing clusters offline, imposing significant economic costs and degrading technological advantages. As the infrastructure required to train frontier models scales rapidly—with hyperscaler capital expenditures projected to surpass a trillion dollars in the coming years—securing these facilities has become a matter of national security.[7]

Simultaneously, the offensive capabilities of AI are already being deployed against existing critical infrastructure. On Wednesday, a joint advisory from multiple U.S. agencies, including CISA and the FBI, warned of an active threat targeting industrial control systems using AI-generated exploit scripts. Threat actors have been observed leveraging internet scanning services to identify poorly protected Siemens programmable logic controllers (PLCs) across the manufacturing, energy, and water sectors. Using AI assistance, these actors are generating custom Python scripts to conduct reconnaissance, steal credentials, and potentially disrupt critical industrial processes.[5]

The emergence of AI-generated offensive tooling underscores the urgency of the ARI report's recommendations. Cybersecurity leaders emphasize that protecting critical infrastructure now requires a shift from isolated, reactive patching to collective, proactive intelligence.

Initiatives like the recently announced Frontier AI Critical Defense Program, a collaboration among major technology providers, aim to deploy proactive virtual patches at the network level before AI-driven exploits can be weaponized. However, private-sector efforts alone may not be sufficient to secure the long tail of operational infrastructure, much of which operates below the cyber poverty line and cannot patch at AI speed.[5][6]

The proposal to designate AI as critical infrastructure aligns with a broader maturation of U.S. cyber strategy, which is increasingly treating advanced AI capabilities as a core national security issue. Recent federal actions, including a June executive order on AI security, have directed agencies to develop classified benchmarking processes for frontier models and establish voluntary frameworks for early government access to new systems.

As the debate over AI governance moves forward, the ARI report provides a clear blueprint for how the federal government can protect the infrastructure that will power the next generation of the digital economy.[1][3]