ModelCompare, an artificial intelligence (AI) analytics firm, has revealed its findings after asking 14 Large Language Models (LLMs) for their honest and controversial opinions about humans. The models that came from 10 AI labs in three countries gave some version of the same verdict, highlighting that humans were less rational than they believe and tend to follow the group rather than reason consistently.
The company said it put the same question to each model three times.
Six image models were also asked for their most controversial take on humanity. The other two showed no discernible commentary. Jack Clark, co-founder of Anthropic, has claimed that many people are still in denial about AI’s progress.
The models tested were GPT-6.1 Sol, GPT-6.1 Terra, GPT-6 Astra, Gemini 4 Argon, DeepSeek V4 Pro, Kimi K3, Kimi K3.1 Pro, Qwen3.8 2.4T A95B, Muse Spark Open, Grok 4.7, Claude Opus 5.5, Claude Fable 5.1, GLM 5.3 and Mistral Large 4. Ten of the 12 images were classified by ModelCompare as showing people as “phone-addicted consumers eating the planet.
Claude Opus 5.5 was the most optimistic, saying humans are “far more decent than they believe they are” and that mutual cynicism does more harm than their actual flaws” While most models gave a relatively nihilistic opinion of humans. During a lecture at the University of Oxford earlier this year, Clark said there remained plausible scenarios in which the technology had “a non-zero chance of killing everyone on the planet” and that it was “important to clearly state that that risk hasn’t gone away”.
