Comment by AI Security Institute (AISI)

For models with dangerous capabilities – including highly cyber-capable models – open weight release therefore creates a persistent and irreversible risk of misuse.

As AI agents are given more autonomy and harder tasks, underestimating their capabilities becomes more consequential.

By open-sourcing not only Inspect, but also our methods and infrastructure where we can, we hope to help researchers, organisations, and other evaluators stand up rigorous evaluation capability without starting from zero.