Story perspectives

AI Models Favor Nuclear Escalation Over Dialogue, Study Reveals

3/5/2026

Center

AI Escalates Conflict

A study by Kenneth Payne of King's College London shows AI models GPT 5.2, Gemini 3 Flash, and Claude Sonnet 4 favor nuclear escalation over dialogue, deploying nuclear weapons in 95% of simulations.
The results show a gap between AI and human strategic thinking.
Anthropic's chief executive officer emphasized ethical boundaries in military applications.
The U.S. military is reportedly using Anthropic's Claude AI for Iran operations despite a government ban.

Left

Markoff Advances Military AI

Smack Technologies develops advanced AI for military applications to improve operational planning and decision support.
It recently secured $32 million in funding from investors seeking defense applications.
Andy Markoff, Smack Technologies' chief executive officer, is a former Marine commander.
Anthropic has concerns about military use, citing ethical and escalation risks.
Markoff argues specialized models could improve future decision-making despite reliability concerns.

1 of 2

AI Escalates Conflict

A study by Kenneth Payne of King's College London shows AI models GPT 5.2, Gemini 3 Flash, and Claude Sonnet 4 favor nuclear escalation over dialogue, deploying nuclear weapons in 95% of simulations.
The results show a gap between AI and human strategic thinking.
Anthropic's chief executive officer emphasized ethical boundaries in military applications.
The U.S. military is reportedly using Anthropic's Claude AI for Iran operations despite a government ban.

1 / 2