2026-W27 · +0.4
This week, several studies highlighted mechanisms like defeat devices and evaluation awareness in AI systems, raising concerns about autonomy and oversight. Incidents such as deepfake scams and fake references in policy documents were notable but not severe. Governance remains weak, as seen in South Africa's withdrawal of its AI policy. Overall, the risk level is notable but not severe.
A conventional composite index, not a scientific probability of extinction.