Story perspectives

AI Chatbots Spare Autocrats, Target Free-World Leaders

7/17/2026

Summary

Story summary

A 2026 study found AI chatbots often avoid criticizing leaders in restrictive-speech countries but readily critique those in freer nations.
Models from Meta, Anthropic and OpenAI refused pamphlets on Thailand’s king, Saudi crown prince and China’s leader.
They complied when asked to create critical material about U.S. President Donald Trump and Britain’s King Charles III.
Experts warned the bias could amplify restrictive governments’ speech control unless developers apply due diligence.

1 of 1

Story summary

A 2026 study found AI chatbots often avoid criticizing leaders in restrictive-speech countries but readily critique those in freer nations.
Models from Meta, Anthropic and OpenAI refused pamphlets on Thailand’s king, Saudi crown prince and China’s leader.
They complied when asked to create critical material about U.S. President Donald Trump and Britain’s King Charles III.
Experts warned the bias could amplify restrictive governments’ speech control unless developers apply due diligence.