Bio Risk

METR

Model Evaluation and Threat Research conducts dangerous capability evaluations for frontier AI models, testing for autonomous replication, cybersec...

Bioweapons Risk

AI-assisted biological weapon development represents one of the most severe near-term AI risks. In 2025, both OpenAI and Anthropic activated elevat...

Misuse Potential

AI Misuse Risk Cruxes

Key uncertainties that determine views on AI misuse risks, including capability uplift (30-45% significant vs 35-45% modest), offense-defense balan...

Johns Hopkins Center for Health Security

Independent nonprofit research organization focused on preventing and preparing for epidemics, pandemics, and biological threats, with significant ...