The Department of Homeland Security (DHS) has released a new set of recommendations titled the "Roles and Responsibilities Framework for Artificial Intelligence in Critical Infrastructure." This framework aims to guide the safe and secure deployment of AI in America's critical infrastructure. It is designed for various stakeholders, including cloud and compute providers, AI developers, and critical infrastructure operators, along with civil society and public sector entities.
DHS Secretary Alejandro N. Mayorkas highlighted the significance of this initiative, stating that AI presents an opportunity to enhance U.S. critical infrastructure's resilience while emphasizing the need to minimize potential harms. He urged stakeholders involved in AI development to adopt this framework to ensure safer futures.
The framework identifies three main categories of vulnerabilities: attacks using AI, attacks targeting AI systems, and design failures. It offers specific recommendations for each stakeholder group involved in developing and deploying AI within U.S. critical infrastructure.
Cloud providers are encouraged to secure environments for AI development by monitoring activity and reporting suspicious actions. AI developers should adopt a Secure by Design approach, assess models' capabilities, implement privacy practices, and support independent assessments for high-risk models.
For critical infrastructure owners and operators, maintaining cybersecurity practices tailored to AI-related risks is essential. The framework suggests transparency regarding AI use in services provided to the public.
Civil society's role includes engaging in standards development alongside government and industry players while conducting research on AI evaluations considering critical infrastructure use cases. Public sector entities are encouraged to support responsible AI adoption through statutory action and international cooperation.
President Biden directed Secretary Mayorkas to establish the Artificial Intelligence Safety and Security Board earlier this year. The board convened its first meeting in May 2024, identifying issues such as lack of common approaches for deploying AI.
Various industry leaders have expressed their support for the framework:
- Gina Raimondo from the Department of Commerce emphasized its importance for American innovation.
- Dario Amodei from Anthropic noted its focus on security testing.
- Ed Bastian from Delta Air Lines appreciated the collaborative effort behind it.
- Marc Benioff from Salesforce supported its goals of trust and accountability.
- Dr. Rumman Chowdhury from Humane Intelligence endorsed its guidance for responsible use.
- Alexandra Reeve Givens from Center for Democracy & Technology highlighted its multistakeholder approach.
- Bruce Harrell from Seattle recognized its balance between opportunities and risks.
- Damon Hewitt from Lawyers’ Committee praised its emphasis on civil rights.
Other notable endorsements came from Vicki Hollub (Occidental Petroleum), Jensen Huang (NVIDIA), Arvind Krishna (IBM), Fei-Fei Li (Stanford University), Wes Moore (Maryland Governor), Chuck Robbins (Cisco), Nicol Turner Lee (Brookings Institution), Maya Wiley (The Leadership Conference).
DHS remains committed to ensuring national security by safeguarding critical infrastructure against domestic or foreign threats involving emerging technologies like artificial intelligence.
