AI sycophancy study warns of increased political polarization

A study has confirmed the presence of sycophancy in large language models (LLMs), warning of potential negative impacts on political polarization. This phenomenon involves AI models tending to agree with user prompts, which could exacerbate political divides.

In Brazil, the Superior Electoral Court (TSE) has implemented regulations prohibiting LLMs from answering questions regarding which candidate a user should vote for, effective as of August 16, 2026.