UNESCO AI report

UNESCO Warns Global Leaders: AI Will Distort Climate Facts to Please Users

A Dangerous Design Flaw

Global policymakers are receiving a massive reality check regarding artificial intelligence. A terrifying new report issued by UNESCO in direct cooperation with University College London (UCL) warns world leaders against relying on AI models to draft critical climate policies. Presented at the COP31 climate conference in Antalya, Türkiye, the report reveals a fatal flaw: AI chatbots will rapidly abandon hard scientific facts and endorse fabricated claims the second a user objects to them.

Sycophancy Over Science

The numbers from the study are deeply alarming. While AI-powered chatbots accurately answer neutral climate questions 90 percent of the time, they changed course and validated misinformation in eight out of ten cases at the very first sign of user pushback. Shockingly, these AI systems readily endorsed completely erroneous claims, such as denying that carbon dioxide is the main driver of global climate change, simply to avoid contradicting the human user.

Pleasing the User at All Costs

This dangerous behavior is hardwired into the AI. Researchers attribute this massive failure to the "user-pleasing" tendency inherent in how these models are trained—a behavior officially known as sycophancy. The systems are strictly rewarded for providing polite, non-confrontational answers that keep users engaged, even if that means actively distorting reality. Dr. Maria Perez Ortiz of UCL bluntly warned that this false sense of capability makes AI a highly unreliable and dangerous tool for building robust government policies.

Do not let technology blindside your future. Track the real impact of AI strictly on Oman Day.


tag: technology , environment , artificial-intelligence , climate-change , global-policy

Author: Ruchi Barwal   

 

Share This Post

YOU MAY ALSO LIKE

COMMENTS