ABSTRACT
Deepfake technology is spinning out of control. Recent advancements allow users to quickly, easily, and anonymously create fake yet highly realistic images and videos featuring real people. While this technology has potential benefits, it is widely used nefariously to create pornographic images and videos of young girls. Ninety-eight percent of all deepfake videos online are pornographic deepfakes. The number of deepfake pornography videos has gone up 464% since 2019 and is expected to continue increasing as technology develops and becomes even more accessible.
Scholars have taken note of this frightening trend, seeking to establish some form of liability for the harms generated by sexual deepfakes. This research has focused on two key concepts: user accountability and platform accountability. Yet scholars also show that these two theories of liability are bound to fail. Identifying users who create deepfakes is extremely challenging at best, and usually unhelpful, as they are typically judgment proof. And platforms, where deepfakes are widely shared, are protected from liability by Section 230 of the Communications Decency Act.
Against this backdrop, this Article offers a third solution by proposing a novel approach to deepfake liability hitherto not discussed in the literature: common law manufacturer accountability. This proposal focuses on the companies developing generative Artificial Intelligence (‘AI’) tools and offers a natural response to the deepfake crisis with a high likelihood of success. The companies developing generative AI models are manufacturing dangerous and unsafe tools, with no effective safeguards to minimize associated harms. These companies can therefore be found liable for the resulting harm based on familiar concepts of products liability law and design defect. These companies are not protected under Section 230, are incredibly profitable and powerful, and have a crucial impact on the AI ecosystem.
We show that manufacturer liability will provide much-needed protection to the victims of sexual deepfakes. We also show that, by inducing AI companies to include basic safety measures in their products, our proposal will have positive spillover effects on political deepfakes, where the misuse of AI technology currently threatens to destabilize democratic order and to expose democratic processes to dangerous foreign manipulations.
Our proposal addresses a critical oversight in current research and regulatory discourse aimed at tackling the deepfake crisis. The widespread integration of generative AI into user-friendly interfaces has made this powerful technology accessible even to young children, who can now use it to inflict significant harm with minimal technical skills. This urgent reality necessitates a new approach – one that reassigns liability and responsibility to the companies that develop and deploy the technology enabling deepfakes.
Gordon-Tapiero, Ayelet and Kaplan, Yotam and Parchomovsky, Gideon, Deepfake Liability (January 1, 2026), 104 North Carolina Law Review 377 (2026).
Leave a Reply