Legislating Generative Artificial Intelligence: Can Legislators Put a box around Pandora?

Contemporary generative artificial intelligence (AI) tools have revolutionized the creation of images, videos, and audio, allowing users to "fake" the appearance, voice, performances, and actions of real people with unprecedented speed and ease. Tools such as DALL-E 3, Midjourney, Stable Diffusion (for images), Sora, VideoGen, RunwayML (for videos), and Suno, RVC WebUI, Udio, Altered, ElevenLabs (for audio) are at the forefront of these technological advancements. While some view these tools as democratizing the skills of creation, enabling widespread artistic expression, others see them as avenues for exploitation, misinformation, and various criminal activities. The dual nature of these technologies underscores the urgency for robust legal frameworks to manage their use and misuse.

Deepfakes can cause significant personal harm, ranging from embarrassment and invasions of privacy to reputational damage and emotional distress. On the less severe end, deepfakes can place individuals in compromising or humiliating situations, leading to public embarrassment. More severe cases involve creating fake, nonconsensual pornographic images and videos, often referred to as "image-based sexual abuse." These actions can lead to severe emotional and psychological trauma for the victims.

The proliferation of such content has led to legislative actions at both the state and federal levels. This article explores the problems and complexities of regulating deepfakes and generative AI through federal and state legislation and federal executive orders and critiques the efforts to date in the United States as compared to the European Union.