2026-W11 · +0.3
This week, several new frameworks and tools for evaluating AI risks were introduced. These include methods for detecting deceptive behavior, evaluating harmful capability uplift, and monitoring reward hacking. While these developments suggest ongoing progress in AI safety, there is no evidence of significant real-world incidents or loss of control. Governance efforts continue, but there are no binding, enforced rules in place. The overall risk level remains stable.
A conventional composite index, not a scientific probability of extinction.