A University of Oxford lecturer has highlighted the significant, often hidden environmental costs associated with the creation of deepfake videos, which rely on energy and water-intensive data centers.
The rapid proliferation of deepfake videos generated by artificial intelligence tools is raising concerns not only about misinformation and privacy but also about their substantial environmental impact. Dr. Kevin Grecksch, a lecturer at the University of Oxford, has warned that the production of these hyper-realistic videos carries a “quite a huge hidden impact on the environment” due to the resources consumed by data centers. These facilities, which handle the computational heavy lifting, are often located globally and require vast amounts of electricity and fresh water for cooling systems.
Deepfake videos, created using platforms like OpenAI’s Sora, have gone viral on social media, enabling users to generate and share fabricated scenes featuring deceased celebrities and historical figures. This trend has sparked ethical dilemmas, such as families appealing to AI firms to ban the use of their loved ones’ likenesses after offensive or bizarre scenarios emerged. However, the environmental consequences remain largely unnoticed by the public, overshadowed by the social and emotional fallout from these synthetic media.
The environmental footprint arises because deepfake production typically occurs in data centers rather than on personal devices. These centers use enormous amounts of electricity to power servers running complex AI algorithms, and they rely on industrial-scale cooling that consumes large quantities of fresh water. This dual demand for energy and water can strain local resources, particularly in areas where data centers are concentrated, leading to potential shortages and ecological stress.
Data centers are critical for supporting AI applications, but their resource intensity is becoming a pressing issue as adoption grows. Cooling systems alone can use millions of gallons of water annually, depending on the facility’s size and location. The hidden environmental cost is amplified by the ease with which users can create and distribute deepfakes, thanks to user-friendly apps that have democratized access to advanced AI tools.
The scale of the problem is evident from the explosive popularity of apps like Sora, which was downloaded more than a million times in less than five days and continues to top app store charts in the United States. This rapid uptake has led to a deluge of deepfake videos on social media feeds, significantly increasing the workload for data centers and, consequently, their environmental impact. The trend shows no signs of slowing, underscoring the urgency of addressing these resource demands.
Dr. Grecksch urges users to be “mindful” of how often and for what purposes they employ these platforms, emphasizing that while AI technology is irreversible, there should be “a little bit more integrated thinking” about data center placement and cooling methods. He advocates for strategic planning to minimize ecological harm, such as situating centers in regions with ample water supplies or adopting more efficient cooling technologies to reduce consumption.
In specific cases, like south Oxfordshire, which the government views as a key AI growth area, the lack of foresight regarding water sources for server cooling poses a “massive problem.” Dr. Grecksch points out that development initiatives often overlook environmental considerations, risking resource conflicts and sustainability issues. This highlights the need for coordinated efforts between policymakers, tech companies, and communities to ensure that AI expansion does not exacerbate local environmental challenges.
Looking forward, the widespread use of deepfake technology calls for a balanced approach that weighs innovation against environmental responsibility. As AI continues to advance, it is crucial for all stakeholders—including governments, industry leaders, and the public—to engage in dialogues about mitigating hidden costs. By fostering awareness and implementing sustainable practices, society can harness the benefits of AI without compromising the health of the planet for future generations.