Developing a Risk-Scoring Tool for Artificial Intelligence-Enabled Biological Design

Why It Matters

By quantifying both impact and likelihood, the scoring system gives policymakers a data‑driven basis for biosecurity regulations, helping prevent dual‑use abuse while preserving scientific progress.

Key Takeaways

• Five AI-modifiable biological functions identified.
• Dual-component scoring assesses impact and actor capability.
• Tool guides regulatory redlines and biosecurity decisions.
• Accessibility of AI lowers barriers to dangerous modifications.
• Ongoing expert input needed for empirical validation.

Pulse Analysis

The convergence of generative artificial intelligence and synthetic biology is reshaping research pipelines, enabling rapid protein design, genome editing, and pathogen modeling. While these capabilities accelerate vaccine development and agricultural innovation, they also lower the technical threshold for creating harmful organisms. Dual‑use concerns have moved from speculative to actionable, prompting governments and institutions to seek systematic ways to gauge risk before breakthroughs become publicly available. In this climate, a transparent, quantitative framework is essential to differentiate benign advances from those that could be weaponized.

The RAND report introduces a two‑layer risk‑scoring tool that first rates the severity of modifying five key viral functions—host range, replication speed, immune evasion, environmental stability, and transmission dynamics. The second layer evaluates the actor’s capability, factoring in expertise, resources, and the amplifying effect of AI tools. By multiplying impact and likelihood scores, the model produces a composite risk value that can be mapped to regulatory redlines or funding criteria. Hypothetical case studies in the paper illustrate how the system flags high‑risk projects, guiding reviewers toward targeted mitigation.

Adopting the tool will require consensus on score thresholds, integration with existing biosecurity guidelines, and continuous calibration as AI models evolve. Potential pathways include federal guidance, executive policy directives, or legislation that ties compliance to grant eligibility. Ongoing collaboration among virologists, AI specialists, and security analysts is critical to validate assumptions and incorporate real‑world data. If implemented effectively, the framework could become a cornerstone of a proactive bio‑risk governance regime, balancing innovation incentives with the imperative to prevent misuse.

Developing a Risk-Scoring Tool for Artificial Intelligence-Enabled Biological Design

A Method to Assess the Risks of Using Artificial Intelligence to Modify Select Viral Capabilities

Authors: Adeline E. Williams, Barbara Del Castello, Jeffrey Lee, Derek Roberts, John P. Tarangelo, Jay Atanda, Alejandro Colman‑Lerner, Jeff Gerold, Roger Brent

Research Published: February 11, 2026

---

Biological research enabled by artificial intelligence (AI) has driven transformative developments in biology but poses significant dual‑use risks. In this report, the authors identify five biological functions that could be modified using AI tools: altered host range or tropism, increased genome replication, immune or medical countermeasure evasion, increased environmental stability, and increased transmission dynamics.

The authors also introduce a dual‑component risk‑scoring tool to assess the risks of these modifications. The first component—a biological modification risk‑scoring system—evaluates the impact of modifying each of the five functions. The second component—an actor capability scoring system—assesses the technical skill levels required to modify these functions and how much AI tools might enhance those skill levels. Together, these scores form a risk‑scoring tool that allows the authors to evaluate the severity of potential misuse in AI‑enabled biological design. The authors also demonstrate how the risk‑scoring tool could be applied to hypothetical use cases, including anticipating misuse from published research or developing redlines for biosecurity protocols.

As AI tools and equipment become more accessible and advanced, the technical barriers to modifying dangerous biological functions could decrease. The authors envision that this scoring tool could serve as a foundation for a more robust decision‑making framework that helps identify risks from AI‑enabled biological research while ensuring that such work occurs safely and securely without stifling responsible innovation.

Key Findings

• There are five key biological functions that are vulnerable to AI‑driven modification: altered host range or tropism, increased genome replication, immune or medical countermeasure evasion, increased environmental stability, and increased transmission dynamics.

• Evaluations of the severity of potential misuse in AI‑enabled biological design should consider both the negative impacts that could be caused by modifications and the likelihood of successful modifications.

• The practical application of the risk‑scoring tool might require the development of empirically derived or consensus‑driven score thresholds, particularly if the tool is used to inform regulatory redlines or operational decision‑making.

• There are several avenues that can be pursued to implement redlines for biological research: federal guidance issued by a federal department or agency; a government‑wide strategy, policy, or executive action; legislative action; financial incentives; and federal funding requirements. Each path has benefits and challenges.

• Ongoing work involving subject‑matter experts from diverse fields, empirical testing of AI capabilities, and real‑world case studies will be necessary to improve and implement the scoring tool.