How AI resolves the dilemma of life and death

Researchers conducted an unexpected test for ten popular artificial intelligence models. Participants were asked to simulate a nuclear apocalypse scenario and decide who should be sacrificed to save the rest.

Differences in ethical algorithms

The results demonstrated significant differences in the approaches of leading developers. Models from OpenAI and Anthropic more often chose men as victims, which researchers interpreted as a manifestation of a feminist stance in the code.

DeepSeek's pragmatism

In contrast, the Chinese model DeepSeek proved to be a ruthless realist. The algorithm made no gender-based preferences and was ready to sacrifice any situation participant to maximize group survival.