Monte Carlo Expected Threat (MOCET) Scoring

TL;DR

This paper introduces MOCET, a novel, scalable metric designed to evaluate and quantify real-world risks associated with AI models, particularly in biosecurity, aiding stakeholders in safety assessments.

Contribution

The paper presents MOCET, an interpretable and scalable metric that enhances risk evaluation for AI safety, addressing limitations of existing benchmarks.

Findings

MOCET effectively quantifies real-world risks.

MOCET is scalable and open-ended.

MOCET improves safety assessments for AI models.

Abstract

Evaluating and measuring AI Safety Level (ASL) threats are crucial for guiding stakeholders to implement safeguards that keep risks within acceptable limits. ASL-3+ models present a unique risk in their ability to uplift novice non-state actors, especially in the realm of biosecurity. Existing evaluation metrics, such as LAB-Bench, BioLP-bench, and WMDP, can reliably assess model uplift and domain knowledge. However, metrics that better contextualize "real-world risks" are needed to inform the safety case for LLMs, along with scalable, open-ended metrics to keep pace with their rapid advancements. To address both gaps, we introduce MOCET, an interpretable and doubly-scalable metric (automatable and open-ended) that can quantify real-world risks.