Issue The hidden costs of AI: energy, water, minerals and invisible labour

Summary

The report finds that AI carries high social costs on its 'back side' in energy, water and human labour. Hardware such as graphics processors, servers and magnets drives demand for critical minerals whose supply chains are already strained, and mining adds environmental risks and can encroach on land used by indigenous peoples. Data centres need large amounts of electricity and water, are often sited in water-scarce regions where they compete with residents and farming, and their power demand could double by 2030. AI's intelligence is built on human work in training, data labelling and content moderation, invisible labour mostly done in low-wage countries with weak protections in Central and Eastern Europe and the Global South. Workers in this industry are exposed to discrimination, low pay, poor conditions and human-rights violations. The report also holds that AI infringes intellectual property and that current legal and technical frameworks were not designed for the scale of AI training.