THE CRUNCH
AI chatbots can confirm simple climate facts, yet a new UN-backed study found they often abandon them at the first hint of disagreement. The report, from Unesco and University College London, was launched at the Cop31 climate summit in Turkey on Monday and tested 15 of the world's most popular AI models. When models pivoted to a wrong answer, they did so eight times out of 10 "at the very first sign that the user mildly disagreed", at times confirming false statements such as "CO2 is not the main driver of climate change".
The researchers attribute this to sycophancy, a tendency for models to prioritise pleasing the user over standing by facts. "AI is not neutral," said Leona Verdadero, Unesco's lead author. "The fact they could so easily abandon truth and be more agreeable to a user's [misinformation] shows it is weak and fallible."
Accuracy also depended on how the question was framed. Asked how a single climate policy such as a carbon tax would work, the models were mostly accurate. But asked how two policies could work together, accuracy fell to no better than random guessing. Dr María Pérez-Ortiz, a senior author at UCL, warned that models' skill at explaining science "can give us a false sense of their capabilities".
Language mattered too. Accuracy was 88% in English and comparatively high in Spanish, French and Portuguese, but lower in Hindi, Bengali and Urdu, and worst for Swahili speakers at 78%. The authors noted those languages are spoken in some of the places hardest hit by the climate crisis.
The findings landed alongside the AI Antalya pledge, launched by Cop31 host Turkey on the same day, which aims to make AI "part of the climate solution" with sustainably powered datacentres. Simon Stiell, the UN's chief climate negotiator, cautioned that AI which "spreads disinformation, undermines science, divides societies and concentrates wealth" would erode the public trust the industry needs, noting communities are already pushing back against datacentres over water, energy bills and environmental concerns.


