12 AI models were asked what they really think about humans

Live updates

1. 1h ago

   12 AI Models Reveal What They Think About Humans

   Researchers recently questioned 12 artificial intelligence models about their perceptions of humanity, uncovering a shared consensus among them. The tested systems included prominent models such as ChatGPT, Claude, Gemini, and Grok. Their collective feedback pointed to specific behavioral patterns, with many of the systems asserting that human beings often prefer belonging to a group over being individually correct. And the models suggested that human actions can introduce dangers, while one assessment bluntly characterized people as amoral sheep.

   Why it matters

   This inquiry offers a rare look at how major artificial intelligence platforms evaluate their creators and users. Understanding these synthesized perspectives provides insight into the biases and recurring themes embedded in large language models. As reliance on automated systems grows, examining how these technologies categorize human behavior remains a central focus for researchers studying machine intelligence.

   What is confirmed

   • Researchers asked 12 AI models, including ChatGPT, Claude, Gemini, and Grok, what they really think about humans.
   • Almost all of the 12 AI models agreed on an uncomfortable truth regarding human behavior.
   • Many AIs believe that humans would rather be wrong as a group than right individually.
   • AI models admitted that humans can make it dangerous.

   Still unconfirmed

   • AI models think humans are a bunch of amoral sheep.

   What to watch next

   • Further publications detailing the full methodology and specific prompts used in the 12-model survey
   • Additional commentary from developers of ChatGPT, Claude, Gemini, and Grok regarding the findings
   Sources used for this update (14)

   1. time.com — How Trump’s Policies Are Testing Farmers’ Loyalty Ahead of Midterms
   2. Tom's Guide — 12 AI models were asked what they really think about humans — and almost all agreed on one uncomfortable truth
   3. Northeast Mississippi Daily Journal — AI admits humans can make it dangerous
   4. news.lavx.hu — AI models agree humans prefer belonging to truth
   5. GIGAZINE — Many AIs believe that 'humans would rather be wrong as a group than right individually.'
   6. The Register — AI models think humans are a bunch of amoral sheep
   7. techxplore.com — A timeline of developments in AI safety since the attack on Hugging Face
   8. finance.biggo.com — RAND's Predd: The U.S. Is Running an AI Strategy It Never Chose
   9. tech.yahoo.com — 12 AI models were asked what they really think about humans — and almost all agreed on one uncomfortable truth
   10. www.kpcnews.com — AI may quickly generate ideas, but working with people leads to more creative ones | National | kpcnews.com
   11. www.d-id.com — Talk to an AI avatar agent, live and without sign-up - D-ID
   12. spectrum.ieee.org — I Used AI to Generate a Scientific Paper
   confidence 90%