Commentary. The climate talks in Antalya have put artificial intelligence in an unfamiliar role: both a possible tool for climate action and a rapidly growing source of environmental pressure. That dual status should shape the negotiations’ outcome. Governments should welcome useful AI, but refuse to treat optimistic pledges as a substitute for measurable rules.

The United Nations reports that COP31 discussions remain divided over climate finance, adaptation, fossil fuels and the protection of workers during the energy transition. At the same time, negotiators are showcasing initiatives linking AI with greener industry and more accessible technology. A voluntary Antalya Artificial Intelligence Pledge had attracted 80 endorsements and 271 commitments by Nov. 16, according to the UN.

Those commitments are worthwhile. AI can improve weather forecasting, help electricity grids balance renewable power and make climate information easier to interpret. For poorer countries, access to reliable data and efficient digital infrastructure could strengthen disaster preparation and support decisions that are currently made with incomplete information.

But the attraction of a technological solution can become a political escape hatch. Better models cannot replace flood defenses, public-health systems or money for countries already suffering climate damage. Nor can they settle the central dispute at COP31: who pays, how much, and how quickly the world moves away from fossil fuels.

The case for guardrails

The environmental cost of AI is no longer an afterthought. Data centers require electricity, cooling and water; the supply chain for computing equipment requires land, minerals and industrial capacity. The precise impact varies by model and location, but that uncertainty is itself an argument for disclosure rather than complacency.

A separate COP31 initiative, launched by the UN, UNESCO and Brazil, calls on AI companies to assess and publish the effect of their systems on climate-information integrity. It also urges disclosure of energy use and carbon, water and land footprints. These proposals address a second danger: AI-generated climate content can amplify falsehoods, obscure scientific uncertainty or make fabricated claims appear authoritative.

Our view is straightforward: if companies want public trust, they should disclose the resources their systems consume and explain how they measure those impacts. Governments should require comparable reporting for large systems, set efficiency standards where evidence supports them, and ensure that public institutions can identify synthetic or manipulated climate information.

Critics will reasonably warn that heavy regulation could slow innovation, raise costs and deny developing countries access to tools that wealthier states already use. They are right that climate policy must not become a pretext for locking poorer countries out of technological progress. Rules designed in Washington, Brussels or other rich capitals can also impose burdens that ignore local energy systems and development needs.

That is why the best approach is not a blanket ban or a one-size-fits-all ceiling. It is a baseline of transparency, independent measurement and access to technical assistance. Smaller developers and poorer countries need support to comply. Public procurement can reward efficient systems without requiring every experiment to meet the standards of a multinational data-center operator.

Promises need consequences

Voluntary commitments have a role: they can move faster than treaties and test practical standards. But they are credible only when progress is reported, comparable and open to scrutiny. The history of climate diplomacy offers ample evidence that declarations without implementation can create the appearance of movement while emissions continue rising.

COP31 should therefore connect AI initiatives to existing climate accountability. Each commitment should identify a baseline, a deadline and an independently verifiable indicator. Claims about emissions reductions should distinguish between genuine reductions and avoided emissions, while claims about climate-information quality should disclose methods and limitations.

There is a temptation to frame this as a choice between innovation and regulation. It is not. Rules that reveal costs and prevent deception can make useful innovation more durable. The real choice is between an AI transition governed in the public interest and one shaped mainly by the incentives of the companies building it.

AI may help the world adapt to a changing climate. It cannot excuse governments from financing adaptation, cutting pollution or telling citizens the truth about the scale of the crisis. COP31 should leave Antalya with fewer slogans, stronger disclosure and a clear principle: technology is climate policy only when its benefits are demonstrable, its costs are visible and its promises are enforceable.

Sources