Artificial Intelligence strains power grids and outpaces regulation, report says

A recent International Energy Agency estimate projects that data centres powering artificial intelligence could account for 3‑4% of global electricity consumption within this decade. The rapid, synchronized training of AI models creates short‑term, high‑intensity loads that are difficult for grid operators to predict, while distributed inference adds further uncertainty about when and where demand will peak. These fluctuations challenge traditional forecasting, reserve planning and frequency‑control mechanisms, prompting utilities to adopt batteries, fast‑acting condensers and other mitigation technologies.

At the same time, a first‑stage report from the International Independent Council on AI warns that the pace of AI development is surpassing most nations' ability to regulate it. The council notes that AI is already achieving expert‑level reasoning in mathematics and science, accelerating drug and vaccine research, but also being used to generate disinformation, facilitate fraud and potentially weaponise biotechnologies. Governance frameworks remain fragmented, with many countries lacking the capacity to assess advanced AI systems, increasing the risk of loss of human control and unsafe deployments.