DHS releases Roles and Responsibilities Framework for AI in Critical Infrastructure

On November 14, 2024, the U.S. Department of Homeland Security published the Roles and Responsibilities Framework for Artificial Intelligence in Critical Infrastructure, a voluntary set of recommendations for organizations across the AI supply chain. DHS describes it as a first-of-its-kind collaboration with industry and civil society intended to advance the responsible use of AI in the country's critical infrastructure.

What it is

The document was produced with input from the Artificial Intelligence Safety and Security Board, a public-private advisory committee that DHS established.

That board was set up in response to the prior administration's 2023 executive order on the development and use of AI.

The framework covers the same 16 sectors of the economy that CISA has defined as critical infrastructure, including communications, energy and financial services.

What it requires

The framework identifies three categories of AI-related safety and security vulnerabilities across critical infrastructure: attacks using AI, attacks targeting AI systems, and AI design and implementation failures.

It assigns voluntary recommendations across five stakeholder groups covering five responsibility areas: securing environments, driving responsible model and system design, implementing data governance, ensuring safe and secure deployment, and monitoring performance and impact.

For cloud and compute infrastructure providers, the framework recommends establishing vulnerability management methods to scan, or enable customers to scan, infrastructure for AI-related threats.

Critical infrastructure owners and operators are urged to incorporate appropriate human involvement in consequential decisions and provide meaningful transparency about their use of AI.

While the framework sets out detailed practices, it does not carry legal force in the way the EU AI Act's high-risk rules do.

What security leaders should do

Security teams operating in any of the 16 CISA-defined sectors should map their AI deployments to whichever of the five stakeholder roles they occupy and compare current practices against the corresponding recommendations. Because the framework is voluntary and references other technical resources for implementation, treat it as a gap-analysis checklist rather than a compliance mandate, and pair it with existing risk-management work such as the NIST AI Risk Management Framework. Leaders should also confirm whether their organization's sector already faces incident reporting obligations from CISA, since the framework's scope tracks those same critical infrastructure sector definitions.