Risk Index Observatory (RIO) for AI Biological Tools

What is the issue?

Artificial intelligence (AI) is dramatically transforming the life sciences, opening new frontiers in drug discovery, synthetic biology and bioengineering. However, the same advanced AI tools that accelerate research and innovation, sometimes known as biological tools (BTs), can also be misused for harmful purposes. For example, AI BTs that design proteins or predict viral mutations could be used by malicious actors to engineer immune-evasive or pandemic-capable pathogens.

Prior RAND work, in collaboration with the Centre for Long-Term Resilience (CLTR), created a Global Risk Index evaluating advanced AI-enabled biological tools for dual-use potential and dangerous capabilities. Among 57 state-of-the-art tools, 13 tools were designated ‘Red’, indicating high concern, with several enabling functions such as predicting mutations that may confer immune evasion in pandemic pathogens. A further 15 tools indexed as ‘Amber’, warranting a case-by-case follow-up assessment.

While existing assessments have provided valuable data on the landscape of AI BTs, they are mostly static and descriptive, offering limited structured, real-time metrics for policymakers seeking to assess and manage emerging misuse risks. Moreover, oversight of BTs remains fragmented across countries and institutions. There is no dedicated global observatory to monitor the rapid evolution, increasing accessibility, and proliferation of AI-enabled BTs, or to provide tailored risk assessments, early-warning signals and practical guidance for both policymakers and developers.

How are we helping?

Building on the Global Risk Index approach, RAND is developing the Risk Index Observatory (RIO): a continuously updated, automation-assisted platform for assessing misuse capabilities of AI BTs. RIO will track and assess new tools as they emerge, translating these insights into dynamic dashboards, early-warning signals, and practical guidance for developers. The observatory will be available as a public website featuring high-level summaries and visualisations, with more detailed dashboards and assessments gated for trusted government audiences.

Eventually, RIO will incorporate automated web searches to detect novel AI BTs in near-real time. A semi-automated triage pipeline, powered by a domain-specific language model, will parse technical descriptions and support preliminary risk screening and prioritisation for expert review. Tools will be systematically assessed for their capabilities, accessibility and dual-use or misuse potential. Assessments will identify tools, sectors, and geographic clusters that warrant heightened scrutiny, supporting proportionate regulation and more efficient oversight. Decision makers will benefit from early alerts on emerging high-risk capabilities, reducing strategic surprise and enabling pre-emptive action.

The Observatory also includes a second stream of work that focuses on practical resources for developers of AI BTs. We will develop an ‘RIO calculator’ that will allow AI BT creators to enter model characteristics and deployment choices, generating a calibrated self-assessment of risk and signposting voluntary safeguards or relevant risk-management resources. This resource will support developers in understanding and mitigating risks as the field advances, without functioning as a safety certificate.