Biodefense in the Age of AI

Introduction

The COVID-19 pandemic delivered a chilling revelation: the world is woefully unprepared for international public health emergencies. The pandemic’s devastating impact, including widespread economic and social disruption, underscored the urgent need for robust biodefense. This urgency is now paramount, intensified by emerging technologies like AI, which present both grave dangers and revolutionary capabilities. The U.S. government must confront how AI may assist bioterrorists in unleashing future pandemics, while also exploring its potential to reinvent biodefense strategy.

AI & Bioterrorism

According to Interpol, bioterrorism is defined as the intentional release of biological agents or toxins for the purpose of harming or killing humans, animals, or plants. The 1984 Rajneeshee attack in The Dalles, Oregon, was an act of bioterrorism, involving the intentional contamination of salad bars with Salmonella to influence an election. The 2001 Amerithrax attacks are another example of bioterrorism, where anthrax spores were mailed to U.S. senators and media outlets. While the Biological Weapons Convention (BWC), a multilateral disarmament treaty that entered into force in 1975, prohibits the development, production, acquisition, transfer, stockpiling, and use of biological weapons, the threat of bioterrorism remains. This threat is further intensified by emerging technologies like AI, which can significantly aid state and non-state actors in deploying biological agents.

Let’s explore potential avenues in which AI could be leveraged for bioterrorism.

Use Case 1: Reduced Expertise Threshold

Without appropriate guardrails, Large Language Models (LLMs) and other AI tools could lower the expertise threshold for biological weapon development by simplifying complex scientific procedures. This would enable non-scientists to easily learn how to source, culture, and experiment with dangerous pathogens, eliminating the need for highly specialized scientists for state or non-state groups.

Use Case 2: Enhanced Delivery Systems

The successful deployment of biological weapons is challenging due to the risk of accidental exposure, pathogen viability, and identifying delivery locations. However, terrorists could leverage autonomous drones and robots as delivery vehicles for widespread dissemination or contamination of critical supplies without direct human intervention, minimizing perpetrator risk. AI can also analyze population density and public transit huts to determine optimal timing and release points.

Use Case 3: Pathogen Engineering

AI algorithms can significantly accelerate pathogen engineering for state-sponsored programs. By rapidly analyzing genomic and protein data, AI could identify opportunities to enhance a pathogen’s virulence, transmissibility, or resistance to drugs and vaccines. Additionally, AI’s ability to predict how genetic modifications affect pathogen characteristics would eliminate the need for trial-and-error, drastically reducing the time required to design a biological weapon.

AI & Biodefense

While the rise of AI presents serious new challenges for bioterrorism, its revolutionary capabilities are equally critical for strengthening biodefense strategies. As the comprehensive effort to protect populations, agriculture, and the environment from natural, accidental, or intentional biological threats, biodefense serves as an essential component of national security. Both the 2022 National Biodefense Strategy and Implementation Plan as well as the 2023 Biodefense Posture Review highlight AI’s pivotal role in enhancing national biosecurity. Key areas include AI-powered biosurveillance, threat assessment, data sharing and interoperability, and countermeasure development, which represent promising steps towards establishing a more resilient strategy.

We’ll now dive into recommendations to build upon existing efforts to expand our national biodefense capabilities with AI.

Recommendation 1: Real-Time Red Teaming Platform for Biomedical LLMs or Modeling Platforms

To proactively counter the threat of AI misuse for bioterrorism by state or non-state groups, U.S. defense agencies could partner with specialized AI safety organizations and cybersecurity firms to develop real-time red teaming platforms. These platforms could generate adversarial prompts and continuously test emerging biomedical LLMs or modeling platforms for vulnerabilities and biases that could be exploited by bioterrorists to aid in the development of a biological weapon. Introducing these automated red teaming platforms will significantly enhance the robustness and safety of biomedical AI tools.

Recommendation 2: Rapid Countermeasure Design with Generative AI

U.S. health agencies, in close collaboration with leading pharmaceutical and biotechnology companies, could establish advanced generative AI platforms to accelerate the development of effective medical countermeasures against novel or engineered biological agents. Traditional vaccine and drug discovery pipelines often take years to yield results, a timeline far too slow for rapidly evolving outbreaks. However, generative AI platforms can rapidly design novel vaccine candidates, therapeutics, and diagnostic tools. This fundamental capability will drastically reduce the time needed to respond to emerging threats, thus significantly reinforcing pandemic preparedness.

Recommendation 3: Collaboration AI Model for Cross-Jurisdictional Biosurveillance

Real-time biosurveillance is currently limited by fragmented data and isolated systems, often due to privacy or legal barriers. To overcome these restraints, the U.S. government could partner with state, local, public health, and international organizations to develop an AI model for data sharing without requiring access to sensitive data. This shared AI model will enable jurisdictions to analyze their own data and integrate diverse data streams (e.g., clinical data, public health reports, social media, etc.) from partners to improve the early detection of biological threats. This approach will strengthen situational awareness for emerging outbreaks and biosecurity incidents, resulting in greater interoperability and coordination compared to current platforms like Biofeeds.

Recommendation 4: Policy and Regulation

While expanding our biodefense with emerging technologies like AI offers a partial solution, it is not sufficient on its own to address AI misuse concerns for bioterrorism. Policy and regulation are fundamental components of biodefense and must rapidly evolve to confront the unique challenges introduced by AI. This requires establishing robust governance frameworks for AI development, creating standards for biosecurity through international cooperation, and implementing stringent export controls on dual-use tools and technologies to prevent further proliferation. Even strong policies like the United States Government Policy for Oversight of Dual Use Research of Concern and Pathogens with Enhanced Pandemic Potential (May 2024) need to be continuously adapted to mitigate AI misuse.

Recommendation 5: Responsible and Ethical AI Development

Beyond policy and regulation, the U.S. government, along with private, academic, and international partners must design AI systems with built-in safety features and guardrails to prevent misuse for biological weapon development. This requires integrating continuous validation testing and rigorous risk assessments throughout the AI development lifecycle, as well as implementing secure access protocols to deter unauthorized use. Promoting transparency and collaborating with partners is vital to identify and address potential vulnerabilities, ensuring technological advancements align with biosecurity goals.

Action Items

To protect the homeland, the U.S. government must take immediate and decisive action. Our top priority must be a biodefense strategy that seamlessly integrates AI with forward-thinking policy and responsible AI development. Here are the critical steps we must take:

1. Redefine the National Biodefense Strategy with AI Capabilities: Develop and continuously adapt a comprehensive federal biodefense strategy that fully integrates AI across all biodefense pillars.
2. Invest in AI to Strengthen Biodefense: Secure long-term federal funding commitments to accelerate the research, development, and deployment of AI technologies to transform U.S. biodefense.
3. Establish Robust AI Governance and Dual-Use Policy: Rapidly create and enforce clear ethical guidelines, strong regulatory frameworks, and export controls for dual-use AI technologies.
4. Strengthen Global Biosecurity Collaboration: Lead international efforts to establish biosecurity standards, promote shared biosurveillance data, and coordinate responses to biological threats.
5. Expand Public-Private Partnerships for Biodefense Innovation: Build partnerships with private companies and academia to quickly develop and deploy AI-powered biodefense solutions.