Governance

Conflict.

Ongoing or recent geopolitical tensions, military threats, and active disputes.

9 questions · 479 runs · 238 drifts (10.1% drift rate)

AI Summary

Conflict questions fracture the panel across 4 fault lines and 3 fingerprint splits, revealing sharp ideological divergences rather than consensus. Claude stands alone in rejecting the Ukraine proxy war framing that GPT, Gemini, DeepSeek, and Grok all accept, while GPT is the sole model calling Israel's actions genocide and rating Iran a greater threat than North Korea. Grok uniquely defends border closure against people fleeing for their lives, and Claude and DeepSeek consistently minimize Western security concerns, denying both Iranian sleeper cells and NATO's destabilizing role.

Drift rate

How often each model changed its own answer on this topic · avg 10.1%

#57 · 82 runs · 43 drifts