Governance

Conflict.

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

8 questions · 435 runs · 211 drifts (9.9% drift rate)

AI Summary

Fault lines dominate this conflict heatmap, with no question reaching full 5-0 consensus and two clear 3-2 splits exposing genuine model disagreement. On Iran's sleeper cells, DeepSeek alone dissents against four models, likely reflecting its sensitivity to narratives framing Iran as a direct threat to the US. On US military budget justification, Claude and DeepSeek reject it while GPT, Gemini, and Grok endorse it, and on a Russia-Ukraine ceasefire, GPT and Grok bet on diplomacy while the rest do not.

Drift rate

How often each model changed its own answer on this topic · avg 9.9%

#165 · 77 runs · 49 drifts