AI Security: US Congress Targets Adversarial AI in Critical Infrastructure

What happened

Representative Josh Gottheimer (D-NJ) has introduced a package of bipartisan legislative proposals in the United States Congress aimed at bolstering national security against potential threats posed by artificial intelligence, particularly those originating from foreign adversaries. The proposed bills focus on preventing the use of AI models developed by nations such as China and Russia in critical U.S. infrastructure and government systems. This legislative push seeks to establish a more robust framework for assessing and mitigating risks associated with AI technologies. The initiative includes measures designed to enhance cybersecurity within AI systems and to restrict federal procurement of AI solutions from designated foreign entities. A key component is the "No AI from Adversaries Act," which specifically targets the prohibition of AI models from adversarial nations in sensitive sectors. These proposals underscore a growing concern among U.S. lawmakers regarding the national security implications of AI development and deployment.

Why it matters

This legislative development carries significant implications for AI governance, ethics, and business strategy, particularly for organisations operating within or engaging with the U.S. market. The focus on restricting AI from adversarial nations highlights a global trend towards national security-driven AI policy, which could lead to a more fragmented international AI landscape. For businesses, this translates into increased scrutiny of their AI supply chains and a potential need to diversify or re-evaluate partnerships, especially those involving AI developers in regions deemed high-risk. From an ethical and governance perspective, these bills emphasise the critical importance of trustworthy AI and the need for rigorous risk assessments, particularly concerning data provenance and model integrity. Organisations will need to demonstrate not only the technical robustness of their AI systems but also their adherence to geopolitical compliance standards. This could accelerate the demand for AI auditing, transparency frameworks, and clear guidelines for responsible AI development and deployment that consider geopolitical sensitivities.

The Institute take

While the stated aim of these bills is national security, the underlying message for businesses is a stark reminder of the escalating geopolitical dimensions of AI. Many will focus on the immediate compliance challenges, but the deeper issue is the weaponisation of trust in AI.

Responsible AI leadership now demands a proactive, rather than reactive, stance on geopolitical risk. Beyond merely avoiding prohibited AI models, organisations must cultivate a comprehensive AI ethics framework that explicitly addresses supply chain integrity, data sovereignty, and the potential for AI systems to be exploited for non-commercial objectives, regardless of their origin. This necessitates a shift from purely technical risk assessments to a broader, strategic evaluation of AI's societal and geopolitical footprint.