UN-backed initiative urges AI companies to publish climate assessments and resource use

A new set of climate-AI recommendations asks technology companies to show their work—not just on the information their systems produce, but on the resources they consume. Launched at COP31 on November 13, 2026, the guidance calls for published impact assessments and environmental disclosures. It is a policy appeal, not a new legal requirement.

The Global Initiative for Information Integrity on Climate Change launched the Priorities for Action on AI and Climate Information Integrity in Antalya, Türkiye. The initiative is co-chaired by the United Nations, UNESCO—the UN’s scientific agency—and Brazil. Its guidance addresses what people are told by AI and what they are told about AI itself.

Assess the answers, disclose the footprint

For AI companies, the recommendations pair an assessment duty with a publication duty. They ask developers to examine how their systems affect climate information integrity—whether people can obtain trustworthy information about climate change—and make the findings public.

• Assess information effects: Companies should evaluate how their AI systems affect the integrity of climate information. The focus is on the systems’ effects, rather than simply their ability to generate convincing language.
• Publish the findings: The requested assessments should become public, making disclosure part of the recommendation rather than leaving the exercise entirely inside the company.
• Disclose environmental use: Companies should publish their energy use and carbon, water and land footprints. The requested accounting therefore extends beyond electricity consumption or carbon emissions alone.

The environmental disclosure request responds to a gap identified by the initiative: people are being asked to accept claims about AI’s climate impact without the data needed to evaluate them. The guidance seeks accountability and transparency, rather than declaring the technology either a climate solution or a climate liability.

The question isn't simply whether AI is good or bad for the climate. It’s whether people can rely on what they’re told by AI and about it.

Charlotte Scaddan, Information Integrity Senior Advisor, UN Global Communications

Fluent answers are not necessarily factual

The co-chairs point to a warning from the Independent International Scientific Panel on AI: “producing fluent text is easier than producing factual text”. The UN General Assembly established that panel in 2025. Its warning underpins the initiative’s concern as more people turn to AI for climate information: a readable answer is not, by itself, evidence of accuracy.

UN News highlights a more immediate risk during climate disasters, when AI-generated images and videos can spread faster than official warnings. That concern extends beyond whether a chatbot answers a question correctly. It also involves the speed at which synthetic material can circulate alongside information intended to keep people safe.

A policy role beyond AI developers

The recommendations come from an initiative launched by Brazil, the United Nations and UNESCO at the G20 Leaders’ Summit in Rio de Janeiro in November 2024. It brings governments, international organizations and civil society together to strengthen research and action on climate information integrity.

The new Priorities are grounded in human rights and build on the Belém Declaration on Information Integrity on Climate Change, launched at COP30 in Brazil in 2025. They also draw on the UN Global Principles for Information Integrity, placing the AI recommendations within an existing information-policy framework.

Governments have a specific task in the guidance, too: integrate information integrity into national AI strategies and climate plans. The policy request reaches both sets of plans, rather than treating trustworthy climate information as solely a responsibility for the companies building AI systems.