2026-W14 · +0.3
This week’s dossier supports only small movements in the AI Midnight index. The clearest increases come from research suggesting oversight blind spots and from documented AI-enabled fraud and deepfake abuse. Evidence on dangerous capabilities and acceleration remains mixed, with some safety and detection improvements also reported. Governance looks slightly weaker because one paper argues that a key EU transparency rule may be hard to implement reliably.
A conventional composite index, not a scientific probability of extinction.