Summary

The Automated AI Risk Outlook (AIRO) is a regularly updated dashboard that tracks the probability of catastrophic risk events, as forecast by an ensemble of frontier AI models. The AIRO dashboard includes forecasts of total catastrophic risk, AI-related catastrophes, biorisk, cyber risk, and other AI-related harms. AIRO runs weekly, so forecasts of catastrophic risk update as the world and AI capabilities change.

Automated forecasting is a promising supplement to human elicitation. The most capable AI models now approach superforecaster accuracy on ForecastBench, and they can forecast many outcomes across time horizons and severity thresholds without fatiguing. Each model in the ensemble conducts its own research through an agentic harness, choosing web searches, reading sources, and submitting its probabilities together with a short rationale and the key sources it relied on. AIRO also elicits conditional forecasts—asking models how risks change if frontier AI capabilities advance faster or slower than expected—offering a way to study how capabilities and policy interventions shape catastrophic risk.