AI does not create a new form of nuclear terrorism but meaningfully raises the odds of existing attack paths by lowering technical, logistical, and detection barriers; this elevates the social cost and calls for urgent policy, regulatory, and market responses.

Main Finding

AI does not create a wholly new “fifth face” of nuclear terrorism; rather, it operates as a powerful cross‑cutting enabler that materially raises the feasibility and plausibility of existing nuclear‑terrorism pathways. That elevation in risk calls for a fundamental reassessment of nuclear‑terrorism frameworks and proactive adaptation of counterterrorism strategy and policy.

Key Points

• Reassessment of orthodox typology: The paper reevaluates the widely used “four faces of nuclear terrorism” (fissile material acquisition, improvised nuclear device, radiological dispersal device/dirty bomb, and sabotage/attack on facilities) and concludes AI amplifies each face rather than constituting a separate fifth category.
• AI as an enabler: AI can lower technical and logistical thresholds by:
  • Enhancing intelligence analysis and target identification (faster, more accurate open‑source and signals exploitation).
  • Automating complex design modeling and simulation (reducing expertise/time needed to design devices or bypass safeguards).
  • Facilitating covert procurement and logistics (synthetic identities, automated searching and matching of suppliers, optimized clandestine supply chains).
• Specific scenario amplification: AI makes scenarios previously seen as low‑plausibility—such as sophisticated sabotage of nuclear facilities and effective radiological dispersal—more credible by lowering skill and coordination barriers.
• Challenge to skepticism: Authors argue that practical difficulties often cited by skeptics are eroded as AI reduces the cost, labor, and time required to execute complex tasks, converting some distant hypotheticals into nearer‑term risks.
• Policy imperative: Because AI multiplies existing risk vectors, counterterrorism, nonproliferation, export control, and nuclear security policies must be proactively updated rather than incrementally adjusted.

Data & Methods

• Conceptual and qualitative analysis: The paper uses a literature review of nuclear‑terrorism scholarship, technical descriptions of AI capabilities, and policy literature to map interactions between AI and the four faces of nuclear terrorism.
• Typology reevaluation: Systematic reexamination of the four faces framework to identify where and how AI changes barriers, costs, detection probabilities, and timelines.
• Scenario analysis: Development of illustrative scenarios (e.g., facility sabotage, dirty bomb construction/use) to show plausible mechanisms by which AI could amplify risk.
• Cross‑disciplinary synthesis: Integration of insights from AI technical capabilities, intelligence analysis practices, procurement networks, and nuclear security operations.
• Limitations noted: Largely qualitative and anticipatory; empirical quantification of probabilities and magnitudes of risk is limited by rapid technological change and scarcity of historical analogs.

Implications for AI Economics

• Increased externalities and market failure:
  • Dual‑use nature creates negative externalities not priced into AI products or services; firms lack incentives to internalize global catastrophic risk.
  • Social expected cost of some AI innovations may rise substantially if their facilitation of nuclear terrorism increases the tail risk of mass harm.
• Need for regulation and market interventions:
  • Rationale for targeted regulation (export controls, licensing, end‑use restrictions), mandatory safety certifications, and liability rules to correct incentive misalignment.
  • Possibility of market instruments (taxes on risky capabilities, subsidies for secure‑by‑design development) to steer R&D investment.
• Public spending and procurement:
  • Governments have economic incentives to increase funding for guardrails, secure infrastructure, and detection/mitigation technologies; cost‑benefit analyses should incorporate the raised expected damages from AI-enabled threats.
  • Public procurement can steer private R&D toward safer architectures (conditional contracting, funding for defensive AI).
• Insurance and risk pricing:
  • Insurability of facilities and certain AI services may change; rising tail risks could increase premiums or reduce coverage, altering private-sector behavior.
  • Development of new insurance products tied to certified security practices could create market incentives for compliance.
• R&D allocation and opportunity costs:
  • Increased nonproliferation and security requirements could shift AI R&D away from certain avenues (or increase compliance costs), affecting innovation rates and labor allocation in the AI sector.
• International coordination and governance as an economic good:
  • Global public‑good problem (cross‑border risk externalities) increases the value of international institutions, harmonized standards, and cooperative enforcement to avoid regulatory arbitrage.
• Analytical needs for economists:
  • Quantify the change in expected social cost from AI‑enabled nuclear risks (to inform optimal preventive spending).
  • Model incentives for firms and attackers under new cost/skill conditions to design efficient policy tools (liability, subsidies, export controls).
  • Assess optimal mix of private vs public investment in detection, hardening, and diplomacy given scarcity of resources.

Concluding recommendation: Treat AI‑enabled amplification of nuclear‑terrorism risk as a material economic externality that justifies coordinated policy intervention, directed public funding for safeguards, and economic research to measure and price the heightened risks so that incentives across firms, states, and markets are aligned with global security.