High-Risk AI Models: Military-Grade Security Needed
Securing the Future: Why advanced AI Needs a New Tiered Risk Governance Approach
Table of Contents
The rapid advancement of artificial intelligence presents both unprecedented opportunities adn meaningful national security risks. While much of the conversation focuses on regulating AI development, a more effective approach lies in adopting a tiered risk governance framework - one that mirrors the precautions we already take with nuclear facilities and cyberweapons. This isn’t about stifling innovation; its about strategically co-investing in security to protect the leading edge of American AI while allowing the vast majority of development to flourish.
The Growing Threat Landscape and the Need for a Paradigm Shift
For decades, the United States has benefited from its leadership in technological innovation, especially in areas critical to national security. However, the rise of advanced AI, especially generative AI models, introduces a new level of vulnerability. The potential for misuse – from the creation of complex disinformation campaigns to the development of autonomous weapons systems – is substantial.Furthermore, the systematic theft of U.S. technology by nations like China, as highlighted in recent congressional hearings, underscores the urgency of the situation.
Current approaches to AI security are largely “light-touch,” relying on voluntary guidelines and industry self-regulation. This is insufficient. The stakes are simply too high. We need a shift from treating AI security as a regulatory burden to recognizing it as a strategic imperative. The current environment demands a more robust, proactive, and risk-based approach.
A Tiered Risk Governance Model: Learning from Existing Security Protocols
The proposed solution isn’t to create entirely new regulations, but to adapt existing, prosperous models used for other high-risk technologies.A tiered system would categorize AI systems based on their potential impact.
Tier 1: Low-Risk AI. The vast majority of AI applications – those used for tasks like personalized recommendations or basic data analysis - would remain largely unregulated. This ensures continued innovation and avoids unnecessary bureaucratic hurdles.
Tier 2: Moderate-Risk AI. Systems with the potential for significant societal impact, such as those used in financial modeling or healthcare diagnostics, would be subject to moderate oversight, including openness requirements and independent audits.
Tier 3: High-Risk AI. The most advanced AI systems – those with the potential to pose existential threats, like autonomous weapons or systems controlling critical infrastructure – would be subject to the highest level of scrutiny. this would involve stringent security protocols, government partnerships, and possibly even restricted access.
This framework would require collaboration between a wide range of government agencies,including the Department of Defense,the Department of Energy,the Cybersecurity and Infrastructure Security Agency (CISA),and the intelligence community. Crucially, it would involve a strategic co-investment by the government, strengthening leading AI labs financially through federal partnerships while maintaining a light regulatory touch on the broader AI landscape.
The Advanced AI security readiness Act: A Critical First Step
The recently proposed bipartisan Advanced AI Security Readiness Act represents a crucial first step towards establishing this robust security framework. The bill rightly tasks the National Security Agency’s (NSA) AI Security Center with developing an “AI Security Playbook” addressing vulnerabilities, threat detection, cyber and physical security strategies, and contingency plans for highly sensitive AI systems.
Though, the success of this initiative hinges on close cooperation with the FBI’s Counterintelligence Division, which is responsible for countering espionage targeting AI labs on U.S. soil. Protecting these labs from foreign actors seeking to steal intellectual property or sabotage development is paramount. Passing this bill is a vital down payment on the security infrastructure America needs to navigate the challenges and opportunities presented by advanced AI.
The time to act is now. Treating the strategic stakes of advanced AI with the same seriousness as nuclear facilities and cyberweapons isn’t radical; it’s common sense. By embracing a tiered risk governance model and fostering collaboration between government and industry, we can secure the future of American AI innovation and safeguard our national security.
Jason Ross Arnold is professor and chair of political science at Virginia Commonwealth University, with an affiliated appointment in the Computer Science Department. He is the author of Secrecy in the Sunshine Era: The Promise and Failures of U.S. Open Government Laws (2014), Whistleblowers, Leakers, and Their Networks, from Snowden to Samizdat (2019), and Uncertain Threats: the FBI, the New Left, and Cold War Intelligence (forthcoming, 2025).*
Image: Midjourney
- NZ fitness professionals detail their approaches to physical wellbeing
- Malaysia repatriates 1,476 Burmese nationals on Navy ships
- Why the Social Security COLA Is Announced in October (daybreakwire.com)
- Komjen Herry Heryawan Officially Appointed as New Head of Indonesian National Police Security Maintenance Agency (archyde.com)